diff --git a/docs/plans/counter-asic-3-status.md b/docs/plans/counter-asic-3-status.md
index 1f7b3643a..564cd00fd 100644
--- a/docs/plans/counter-asic-3-status.md
+++ b/docs/plans/counter-asic-3-status.md
@@ -505,6 +505,15 @@ THE 20:2x LINES (BST). The sixty-first landing on master at fc11a091. The AMD an
 
 THE FOUNDER'S STOP (20:21 UK, through main): "STOP all these mess ups." Four tonight: a fix that did not compile announced as cut (777214af); a native AVX-512 gate build rolled as a fleet binary (79 hosts would crash); the workers page never published while he watched; fifteen Mac processes killed by a grep (the fourth kill-by-name breach today). From now, by construction, not by rule: (1) pkill and killall refuse on the Mac (a shim first in PATH, exit 97) and on every box by 20:45 (the build-server lane); (2) a fleet binary comes only from the shipper's cross-build kit at the x86-64-v3 baseline; a gate check scans kits for AVX-512 (the steward, 21:30); a sha is "cut" only when its binary exists on build-1 with its commit string read back; (3) nothing is published from the Mac except the site lane's one deploy script, which records its own pid before it starts; the workers page is the build-server lane's from build-1; (4) a slip is reported once with its cause and the new clock, never re-announced; no lane says "done" for anything it has not read back. Broadcast to every lane in those words at 20:2x; the standing text in CLAUDE.md with this landing.
 
+THE STOP READ BACK, THE ROLL, THE MASTER PIN DIFF (20:2x to 20:3x BST). The sixty-second landing on master at 68e1ea35 (the STOP in CLAUDE.md). By construction, read back: the Mac pkill and killall shims at /opt/homebrew/bin (exit 97; the shipper); 120 of 121 fleet boxes and rented pods shimmed at 20:26 (bps-seed on its next reach), every fleet script killing through /root/fleet/pids (box-dn3.sh, fleet-stop.sh the only stop, box-prover.py finding the SP1 server by the pid owning its socket; a5c934b44); the node lane's scripts carrying no pkill (the one live canary line removed), every pair under /srv/artefacts/200-*/node-lane stamped gate evidence only; the PCs' relay agent refusing any task body that ends a process by name (Stop-Process -Name, taskkill /IM, pkill, killall; exit 77, lab-signing-20, under test); the steward's tools/ci/kit-isa-check.sh (zmm, EVEX opmasks, EVEX-only mnemonics over every executable of a kit; the shipper's /srv/workers/fleet/<sha>-node-lane.tgz only, never the gate pairs), the recorder refusing a "cut" batch without the binary's build-1 path, the commit string equal to the manifest sha and the kit check's clean line, kill-by-name rule 5 (only ^-anchored command patterns, the bracket form, -x, -F or a variable pass), the shim files in the tree at tools/ci/no-kill-shim; the research lane's five relays; the update-return lane's last Mac publishes at 19:44 (the zips from build-1 from now). Two breaches reported once before the shim: the research lane's pkill of its own 20:42 wake-up loop at 20:23 (one sleep loop, nothing else); the hash lane's on box 2 at 19:3x. THE ROLL: the 7cfa422a fleet kit = the build-server lane's cross build on build-1 (/srv/workers/fleet/7cfa422a-node-lane.tgz ea3f2afc, the hive chain's glibc-2.31 pair, "igneumd 2.0.1"; 013986fe / f922b216); roll-kit.sh reads the kit's sha on the Mac then on lp-4090-11 (no AVX-512) the shas, igneumd --version and the miner's version string with a crash-exit check before any restart (its first run refused on igneumd's usage exit code 1, corrected to the version string, one minute); passed at 20:29, the roll from lp-4090-11 now, tens after its four read-backs, the hubs' nodes untouched; the AX162 gate pair removed from every box at 20:09 and never ran; the Mac 2.0.1 entry live (7059a31c) and the HiveOS entry (a16add6a) on the dl host at 20:26; 61 vercel processes alive on the Mac, none ended by name; the pool seat's datacentre leg on lp-4090-12 joined its pool at 20:26:11 (POOL MODE, votes 265 to 273 accepted). The hash lane's ids read back from build-1 at 20:24: the object's pack hl-v6-all.tgz sha256 6df3d4307202fe68b9e846b38c8c7758effd65daca47c65103401f9092a6ddb5, program id 0x4de7b836cc40a4ea, generator 6, program_class v6, class mx8-erad810f22d+sh256x27+state+reg64c+fold+rw (layer 8 on); the candidate hl-v6-all-nowin.tgz 67ceec05d5316141a7c9e250bb192e7d09af11560d90e555fdb71edf9339030c, id 0xbe1d6f48928cef4a; hl-v6-foldrw.tgz f0dc3428 (id 0xd7eba30115d26dd4); hl-v5-nowin.tgz afe6034a (id 0xdace2893e7653830); hl-v6-all-prefix.tgz 8a458bb5 (the same id, its kernel_bound.cu carrying the 384 running-total updates and the 32 prefix loads); the P01 references on build-1 /srv/artefacts/packs/p01-vectors/ hl-v5-win.txt (bd2e862b, pairing with hl-v5-win.tgz e3da3669) and hl-v6-all.txt (bbb757bd), one million lines each; foldrw and all-nowin by 20:45; the tip class-v6 b38c09f45 (two commits over ac86d7910, test-side plus the export's file order; the draw, acceptance and hash b24dfc162's), the suite on build-7 by 20:45 (one slip once: the pack tests reading identity.json). The census hand's slip once: build-6's pool held by the family-gate lane at release class with three more jobs queued, its (c''') read moved to build-4 (class-v6-census-all3 6debbac1a reproducing the pack's id 4de7b836cc40a4ea at generator 6 from its seeds, paired with the 20:24 read-back), the rule, the floor and the attempts census over 256 seeds by 20:50; the window trace at 2^22 on 16 seeds standing from 19:54. The AMD/Intel energy documents landed at e4e0512ce (20:30, 7 files, 622 lines); its registry branch amd-intel-energy-registry caca93c04 (the cell bench:amd-intel-energy on GPU-03, the batch, the map page) unrecorded because the recorder replaces GPU-03's single evidence record (the fleet's 4090 cell) and re-stamps every NOT RUN case's time: the steward's call by 21:00 (a recorder fault either way); the 7600 metered rows about 05:30 (the grid job's cap at 04:51) or 40 minutes after the founder's click. The research lane's 21:00 tip early at 20:23 (counter-asic-4 24883757: the 4 GiB decision FROZEN, the three-row chip line, the window core's approximate figures, F05 corrected on the adversary's placed-core check, the B580 at 11.0 MH/s with its watts unread, nine ECO cases by batch), one slip once (merge markers in the export list, the map page stale), the sha about 20:50. The Master Edition on master at f677914d (20:26); the REV suite (44) and the INT suite (18) with the counting note, REC-01's closure and the two rulings as dated notes by 20:50; F02's bypass check RED on 7cfa422a as the code stands; F03's packaging/release-manifest.json for 2.0.1 on f03-manifest-201 (04d72076), release-manifest-check.sh green, build-from-manifest.sh building kaspad, igneum-pow and igneum-miner from the manifest (the pool not: pool/src at master predating KeyReveal.sig_scheme, the pool-2.0 line landing tonight), the same-work test specified (docs/plans/igneum-2.0-same-work-test.md); P01 BLOCKED until the references (now on build-1). F08's outcome ledger landed b7507dd79 (closed ahead of 23:00); the enforced lane's residuals: V6-10 (the guests rebuilt deterministically at the p22-stage-2 tip, a versioned guest input format, the repinned ids by 23:30, superseding the 17:58 pin), V6-08 (a sub-lane, design 23:00, build 02:00), V6-07 (a sub-lane with the app lane's device coordinator, the 3060 and 4060 rows by 01:00). The node lane: V6-09 a harness cell for 12:00 tomorrow (the minimum validator fed cold valid proofs and expensive-invalid payloads, the in-flight bound, a size cap on proof bytes at the relay as a consensus-safe limit on 2.0.2); I03 folded into the 23:00 native finality plan (authority succession and strip across the window, a voter's restart mid-window, contradictory certificates in both orders, the seed boundary inside a pause, the Sepolia verifier never presenting a recovery certificate as final); the accelerated simulations evidence of rule shape, never operating history. The pool seat: both rented Vast hosts (Haswell) died with Illegal instruction at the class v5 catch-up on the gate build and the hive kit's node alike (the x86-64-v3 class), destroyed at 20:07; the pool on build-2 (its own devnet-4 node on 12424341, 142.132.249.238:4473, TLS pin 39bf9c9d, PPLNS window 2 blocks, fee 1 percent), two CPU members on jobs since 20:30, lp-4090-12 joining; a real pool fault found and fixed on pool-2.0 (no class v5 day-state source on devnet-4, "class v5 needs the execution state"; the miner's provider shape added, state_provider.rs, --exec-rpc defaulting to --evm-rpc); the wire-level chain id 4463 against the node's 4465 by the v0 --network convention (a 2.0 row). Floor lane 3: batch 01 and 02 run on P12's approved grid exactly (eco05.py lines 11 to 14; the cube's 2,592 cells; the 0.06 / 0.12 / 0.25 figures only in the coexistence model's snapshot sections). The adversary lane's slip once: ADV-01's placed 32-lane full core 22:00 to 22:45 (ABC's mapping at 20:25 on 393,000 cells). Lane D's Mac gate reading: the pre-push hook's text and structure checks with the box-side checks over ssh (355 to 414 s), no crate compiled or suite run on the Mac; the default stands. MAIN'S MASTER PIN DIFF dispatched (docs/plans/igneum-2.0-master/pin-diff.md: 48 missing pins, 10 uncovered R1 residuals, 16 disagreements, 12 over-stated statuses); three founder decisions put with defaults: (a) the master's independent commissioned adversary design and second reviewer (p. 3, p. 30, P04) against the in-house rule: default the in-house lanes continue, the R0 to R2 reviewer the reviewer of record, a commissioned outside design held; (b) P11's D5 at ten verified unaffiliated operators, three domains and 30 days: default the in-house exercise labelled "D5 rehearsal", D5 NOT RUN until the public miner ships and the window starts; (c) the master's 17 profiles PROPOSED: default the 18:2x approval stands (P13 deferred); the six-stage sequence (p. 14: 01 Contract through 06 Contender evidence) the programme order, stage 01's exit tonight and tomorrow morning; status discipline: four decisions only (NOT RUN, BLOCKED, FAIL, PASS), kills FAIL, "63 RUNNING" reads "63 in progress, NOT RUN", the steward changing the registry by 21:30.
+THE 20:3x LINES: lane D released build-6 (the width-4 run resumed at 8 cores from its 1,332nd era, 16 cores held and 8 when the predicate row ends; the width-1 control and the subset attack-f8 re-queued in the measure class behind the census hand's v5 leases; the two release-class waiters withdrawn). The hash lane: R1 V6-02 landed on class-v6 (Program::loads_per_hash and bytes_per_hash reporting the EXECUTED schedule, 256 loads and 1,024 bytes of demand per hash under the window against 128 and 512 before it, the drawn count beside and an assert they agree except by the window's factor of two; the reg64 packs re-exporting after the suite, the ids unchanged, the tarball shas named when read back; the served sentences to the site lane); the case-id map for the freeze note (A02: POW-02 and POW-03; A03: POW-03 and ADV-02; A04: POW-02 and POW-04; A08: POW-02 and POW-06; A09: POW-04; the note itself GOV-05); a gap found by the fleet at 20:32: the kit workers' pack reader refused generator 6 ("not a generator version this worker runs (2, 3, 4 or 5)"), fixed on class-v6 (the reader running 6 as 5), the Linux CUDA worker building on build-5, on build-1 by 21:00 when read back; the fleet's F05 pods holding with the ptxas rows (hl-v6-all 100 registers and the prefix form 248 on sm_120, no spill; 93 and 154 on the 4090's sm_89); P01's three v6 packs waiting on the same binary. The enforced lane's V6-07 sub-lane (ad65f923ddf88c324, branch v607-floor-memory) asked the app lane for the device coordinator's interface by 20:55 (default: a memory profile table chosen from the device's free memory at start, the hook named as the next step; rows by 01:00).
+THE CENSUS (c''') READ ON THE FREEZE'S PACK: PASS (20:34 UK, the census hand on build-4). The object: program 4de7b836cc40a4ea, generator 6, class mx8-erad810f22d+sh256x27+state+reg64c+fold+rw, attempt 0, read back from build-1 (sha256 6df3d4307202fe68...), reproduced from its seeds (the Devnet 3 genesis epoch hex, the day hex, era 0, the node1 state) by the harness on b24dfc162's draw (class-v6-census-all3 6debbac1a; the pinned igneum-pow sha256 be3e32dcaa3d958f byte-identical on build-4 and build-6; the class-string path and --program-class v6 the same id). (A) the whole rule with the (c''') floor on the pack's program ACCEPTED: per site over 2^20 evaluations the distinct ratio 0.99994 at the worst site (every site clear of 0.98 and 0.995), the largest 256-item bucket +5.00 sigma, the worst free index bit 3.40 sigma, no site over 6 sigma. (B) the attempts census over 256 chain-shaped seeds under the pack's era and state: 256 of 256 accepted, 0 exhausted at the 256 cap, r 0.129 per candidate, mean attempt 0.15, max 2, (c''') refusing 0 of 294 candidates, the rejections (a') 23, (b) 7, (a) 4, the reg64 liveness rule 3, (c) 1. (C) the window trace at 2^22 on 16 seeds standing from 19:54 on the same instructions (the draw 04442d9ca's; b24dfc162 and 9415e9c89 moving the id recipe and the acceptance's width set-aside, not the draw): the per-site distinct ratio 1.0026 to 1.0027, the item share 1.11 to 1.41x flat against foldrw's 1.10 to 1.37x, paired 1.015 to 1.032x. The record (census-packs.md section 7, the TSVs) on the branch within ten minutes; the registry cell census:class-v6 after.
+THE FIRST FIXED-MINER BLOCK AND P01 PART A (20:3x BST, the fleet): lp-4090-11 (RunPod RTX 4090, EPYC without AVX-512) restarted on the shipper's 2.0.1 kit (ea3f2afc, igneumd 013986fe, miner f922b216, both read back on the box before the restart) at 20:31 on its kept datadir, accepted block 1aa32e5578f6d4b14e4ee214438f0beeba67d7cfb3848ee0e5607975c4b333db at 20:33:33 and 1618b024cc221b1a a third of a second later; zero "amount high word" refusals since; the kept MINE=1 boxes roll in tens from about 20:37 after its five-minute read, then MINE=0, then the 23 peerless boxes on empty datadirs. P01 part A on hl-v5-win PASSED on the three CUDA cards (5090, 4090, 3090) at 20:32: one million vectors each against the hash lane's reference, agree 1,000,000, disagree 0, missing 0, 42 s per card (the JSONs on build-1 /srv/artefacts/p01/p01-20261008-01/); the three v6 packs and F05's bench rows waiting on the generator-6 worker (by 21:00 when read back); the ptxas rows: the prefix form 148 registers over the fold form on sm_120 and 61 on sm_89, no spills. The node lane's (b) 83bb8ea6 placed on build-1 (igneumd 1f53f548, fingerprint 6cdd922a, the accept-fix tree), its canaries next; the last re-cut of (c) on the sha the hash lane names green when read back.
+THE STEWARD's REGISTRY CORRECTIONS (20:37 UK, one landing by 21:00): the decision vocabulary exactly NOT RUN, BLOCKED, FAIL, PASS (RUNNING read as NOT RUN with in_progress_since, 72 rows "in progress, NOT RUN"; P13 DEFERRED read as NOT RUN with the founder's note; any other status refused); the kills read FAIL (POW-03 the connected state, register row 18; POW-07 the mixed FP32, row 12; two experiment cells, the claim impact named); no case PASS beyond FIN-02's simulator half (EVM-01 to 06 NOT RUN with the reference-apps lane's partial records); the REV cases on the master's stable R2-Fxx-Ryy identifiers, INT keeping INT-xx with source R1, the counting note a registry-level note; every evidence record on the p. 139 schema (release identity: commit, lockfile, binary, network object, activation, profile hashes; requirement id; method, never conflated; decision; claim impact; reviewer, empty allowed and never inferred); V6-12 as INT-07's definition and a v6_12 block in the release manifest (signature, lockfile hashes, kernel and host binaries, devices and drivers, the prover patch, memory thresholds, tuner identity) reading BLOCKED until its owners fill it; the four approval layers as four fields on every case, null until a batch sets one; the AMD/Intel lane's block answered by a second evidence record per case (evidence_records keyed by cell, the decision combined, the NOT RUN re-stamp gone); P01's v6 packs BLOCKED on the generator-6 worker.
+THE WORKERS PAGE AND THE KILL GUARD (20:3x to 20:4x BST, the build-server lane): all nine box rows live at the edge (https://dl.igneum.network/fleet-22adafa34bc2/workers.html, workers.json generated 20:24:11, last-modified 20:34:55, build-1 to build-9 with the mac and pcs rows; build-3's row shown though down); the cause of the three-box page: push.mjs hardcoded HOSTS to build-1/2/3 (now enumerating every build-server-N host file, a45d8286 on build-server) and no collector on build-5 to 9 (installed with its 15 s timer); one slip once: publish-fleet's inner Vercel deploy Terminated by the duplicate-deploy cleanup at 20:27, a later deploy carrying the content. The kill guard: pkill and killall refuse with exit 97 (IGNEUM_KILL_OK=1 the override) for the build user on build-1, 2, 4, 5, 6, 7, 8 and 9 (self-test in login and non-login shells on each) and on the Mac (/Users/joshm/.igneum-guard first in PATH; the shipper's /opt/homebrew/bin copy beside it); build-3 on its return; the rented hubs the fleet lane's (shimmed at 20:26). The move-off-the-Mac publish (a small Vercel project from build-1, a uname refusal, .vercelignore) by 21:30.
+
+THE 20:4x LINES (BST). A fault of the coordinator's own, once: the 20:4x line to main said the workers page showed nine boxes at the edge on the build-server lane's publish exit, while main's read at 20:35:47 carried one box (the old single-box collector's file); rule four applies to the coordinator: an edge claim is read from the edge by the one who writes it; read by the coordinator's own curl at 20:45: https://build.igneum.network/workers.json generated 20:45:37 by merge-workers.mjs on igneum-build-1, boxes igneum-build-1 through igneum-build-9, mac, pcs 2. How it works by construction (the build-server lane, 20:43): each box's own collector writes its single-box workers.json served over build-N.igneum.network with CORS (the vhost added on 7/8/9); on build-1 igneum-workers-merge.timer (20 s) runs merge-workers.mjs, curling every box's public feed (no ssh key, no secret), assembling boxes[] and merging the mac and pcs sources; Caddy serves the page, shape.js and the json; the old single-box collector timer on build-1 disabled (the clobberer); nothing touching Vercel or the Mac; the dl fleet-path copy the retired Mac path; the mini to be merged from its relay report (38527a9e). P01 part A complete on the three CUDA cards (the steward's read at 20:41 and the fleet's): on hl-v5-win (e3da3669, id 0x6554474f410f36f3) the kit's CUDA worker and the CPU reference agree on all 1,000,000 nonces on the RTX 5090 (42.5 s, sm_120), 4090 (42.4 s, sm_89) and 3090 (sm_86), zero disagreements, zero missing; the three v6 packs PASS on the generator-6 worker (igneum-worker-cuda-gen6 804a6f7f on build-1 and the five pods; its --check on hl-v6-all "check PASS ... class v6 (state leaves 93); self-test PASS" on sm_120 and sm_89); twelve JSONs on build-1 /srv/artefacts/p01/p01-20261008-01/; the four references with their .sha256 (hl-v5-win.txt bd2e862b, hl-v6-all.txt bbb757bd, hl-v6-foldrw.txt 5f8a30be, hl-v6-all-nowin.txt df1d5477); POW-01 "NOT RUN, in progress" (PASS needs every supported backend: OpenCL on PC 1's RX 7600 the hash lane's, Metal the mini's morning run); the meaning: the CUDA kernels compute the class v5 hash bit for bit across sm_86, sm_89 and sm_120, the first of P01's criteria met in the field. Review B's F05 GPU rows (the fleet, gen6 worker, stock, 250 batches of 2^24, 1 Hz board power): RTX 5090 fold 70.65 MH/s at 437.8 W (6.20 nJ per hash, 100 registers, 16 blocks per SM) against prefix 70.17 MH/s at 500.0 W (7.13 nJ, 248 registers, 8 blocks per SM); RTX 4090 fold 31.31 MH/s at 238.1 W (7.60 nJ, 93 registers) against prefix 31.29 at 231.7 W (7.41 nJ, 154 registers); fingerprints equal (e8f4f328), no spills; the prefix form buys nothing in rate (both bound at the dataset reads) and costs 14 percent more power per hash on the newest card; the GPU side reads against it; the chip side the adversary lane's from the same rows; evidence under /srv/artefacts/tas/54900078/ and /srv/artefacts/tas/si1xc4yhpakk1v/. The roll's first stage closed at 20:42:05 (lp-4090-11's four read-backs: "igneumd 2.0.1", referenceStreamServed true, zero template refusals in 320 s, 186 accepted blocks since 20:33:33 at the stalled chain's low difficulty); hub-1's last 300 EVM blocks after it carry 9 distinct keys, led by lp-4090-12's key through the pool seat's pool (148 of 300) and p1-5090 (131); stage A (the 24 kept mining boxes) rolling in tens since 20:42:06, then the 40 non-mining nodes, then the 23 peerless on empty datadirs. The rule v4 rows on master at c4a2b0088 (the merge of 9ae6c94f, 87 checks): sim/results_v2.md "Rule v4" with the seven real-node runs, v3.mjs's pass line per variant, the batch fin-v4-realnode-20261008T2025Z onto FIN-02 (PASS, the real-node half), FIN-07 (PASS), FIN-08 (RUNNING through the finality-sim cell until tomorrow's combined-boundary rows). The key-succession landing re-shaped: the real pin 58a5aba09 moves elf/'s top pair to afd952e89's and the served pair to prior/, and master's elf/ is what every node kit embeds through include_bytes, so landing it tonight would have every 2.0.1 kit built from master embed the next pair at the top and refuse to start under devnet-4's object (291ee6ae's start check refusing that shape; 5eab958f's acceptance on the 2.0.2 line only; the release-manifest check red for the same reason); what lands is key-succession-pin-tool (2ceb09be9 plus master: the pin tool, elf/prior/ as a copy of the served pair, the manifest without a prior block, the two-pair check; its gate af9491cd), moving no pair; the real pin stays on key-succession-pin on both parent mirrors as the enforced lane's build input and lands with the digest move when main names H. The 2.0.2 line d049b0fa (P22 stage 2 merged; one merge-join brace repaired). The D1 state at 20:38: (a) 3bd95052 six suites green on build-9, the pair on build-1 and build-2 (igneumd 9b0cbeaf, fingerprint a65e4c5a read back from the binary), canaries agreeing on be5f4068 and 0x1171 with the devnet-3 override refused; (b) 83bb8ea6 the same (pair 1f53f548, fingerprint 6cdd922a; one exec test red once on build-9, rerunning); (c) 1a21d1ec six suites green on build-1 including the generator-6 engine arm, pair 90a66c71 on both boxes, canaries the same; the open conditions the hash lane's: the suite green on its final tip 43f1b1581 (master 964cdf6c1 merged into class-v6 for rule 24 over db63e1e36's pack-reader fix; one test-side red in tests/v6fold.rs, a class-name suffix assertion, fixed in the next commit; green by 21:05 when read back), then the last re-cut of (c) (35 minutes, about 21:40), lane D's 6.12 by 22:00, the reg64 packs' re-export shas. The census hand's record at class-v6-census-all3 151edcf99 (section 7: hl-v6-all 4de7b836cc40a4ea at generator 6 PASS; the harness commit 6debbac1a; the TSVs under logs/census-packs/hl-v6-all-g6/ and hl-v6-all-pr/), landing behind the research lane's copy (refused once by rule 26 on the map page at 20:38, master merged in, re-gating, the sha about 21:05). The V6-08 design on the mirror ahead of 23:00 (v608-proving-tasks 4b148a3d1, docs/design/proving-task-protection.md): a claim a lease on the node (300 DAA default, a second claimant refused inside it, the segment reopening at expiry with the holder's late record still accepted, a key with three abandoned leases in an hour deprioritised by 120 DAA, never banned); the fleet prover stopping when the executor is blocked, re-executing, discontinuous, over 64 blocks behind the sink, the DAA stalled 120 s or finality paused 600 s; "smaller" one shard of one block (TASK=auto putting cards under 16 GB on shard tasks); the pool and RPC refusing a proof over the relay's 8 MiB bound before hashing; a preflight of host and export sha256 against a shipped manifest, --mode id against the node's pinned ids, ldd and nvidia-smi; a fourth outcome "abandoned" with the conservation line claims = paid + expired + cancelled + abandoned + active; nothing moving the live digest; the build on both branches by 02:00. The mini: 2.0.1 staged, the install asked at 20:40:09 (the helper quitting the app, swapping the bundle and relaunching). Lane D's 6.11 landing documents-only in its gate; the 6.12 rows on build-6 by 21:40, the landing of 6.11 and 6.12 together by 22:00 with the batch at manifest ac86d7910, method native.
+
 A SHARED-DEVNET FACT FROM THE FLEET (not this lane's, with the shipper and the infra lane): the Hetzner live seed 188.245.5.161:26611 is still on the old override object (digest eada4bda) 1 h 40 min after the 0.3.20 sweep (the fleet never touches Hetzner nodes, so it was outside the sweep); the 0.3.21 wipe canary c22-1 took five digest-mismatch rejects from it; an app with the packaged peers is refused at the seed and syncs through node1 and the hub only, a fresh joiner with only the seed cannot join, the 14 voters and the hub are unaffected; the owner puts the floor file ov16-floor-900000.json (sha 294f1f80) and the c4459193 pin on it. 0.3.21's STAGING (the node lane): the order dry-merges onto 55768f88 with nothing moving to 0.3.22; the late-join fix is 52e96c94 (70e4601e rebased onto 55768f88, exec suite 33 green with both new tests); f067f7c1, b0444f51 and 437f0438 merge clean in order; 2e32d5f6's one conflict (DST_ADDRESS beside pool-finish's DST_BINDING in consensus/core/src/finality.rs) kept both; the live-file digest eada4bda after each (every switch at never); the staging waits on the shipper's sweep-end word; the re-pin held. PC 2 DOWN AGAIN (main, 16:5x UK): the founder takes PC 2 down for cable work (PC 1 back but his desk); both PCs out of the sweep's waves, each updates on its poller on return; no PC job to PC 1; the Windows G1 completed before the outage, nothing reruns. 0.3.21's SECOND GATE LINE on 55768f88 (sha256 279b1b690e854fc9): the ten-minute mixed-version gate beside the 5899f603 pair, 13:37:40Z to 13:47:52Z, SUMMARY PASS (one digest b0afb2ee on five nodes; 223 new and 381 old blocks accepted by the old hub, 0 rejected; counts equal at 319, 486 and 604 through both clean joins and the restart step at 13:45:22Z; no panic); the node lane's two lines on 0.3.21's first candidate complete, in plan 6.9 on ca3-v4-node; the fleet's set on it (the bare-child 12 GB line, the wipe, the kept read, the cases) is the fleet's. 0.3.21's FIRST GATE LINE on 55768f88 (sha256 279b1b690e854fc9, the string read back; pairing igneum-pow 8c728ca3 at byte 5): the digest gate 13:35:41Z to 13:37:19Z SUMMARY PASS (a89be8a7 on both binaries with the peers; db9a85f9 refused, no peer; the live file's eada4bda unmoved); the ten-minute mixed-version gate from 13:37:40Z, line about 13:50Z. The 0.3.21 order as the shipper sent it: 55768f88; f067f7c1 and 70e4601e; b0444f51; 6eb21fc9; db28d331; then the re-pin from 8bdcbdd8 on the coordinator's word; suites between, the digest read after every one; the mirror's release-0.3.20-node back at the pin c4459193, release-0.3.21-node open at 55768f88. THE LATE-JOIN COMMIT (N9's second half, the node lane): 70e4601e on the box mirror as branch proof-hold-fix, from c4459193, two files (igneum/exec/src/proving.rs, protocol/flows/src/v10/proving.rs); the gap was the fetch side on the joiner (the served record ran the native check against the joiner's trailing exec state before anything was stored, the check refused it, the proof was never held, the body rule read "not held" for 20 s and failed the IBD); the fix holds the proof by hash before the checks (the pool entry still needs them) and the serve side says when it holds fewer than asked; the exec suite 32 passed at 13:26Z with the known-failed shape first, the flows check green 13:28Z, igneumd on build-1 at the 0321 worktree path built 13:32Z, sha256 17649eeb2f7d1290, string read back; with the testnet lane (the resume form, B alone); it joins the 0.3.21 staging as its own commit. THE WIPE CANARY ON c19-1, c4459193 (sha 45be9b02d1b002f5, string read back): FORM END rc 0 at 13:50:53Z. Wipe synced 13:35:50Z (57 minutes, inside the 98-minute class); mining 13:36:00Z to 13:47:07Z, 66 mined, 66 accepted, 0 rejected, isSynced true at the tip throughout; the hub holds 41 of its blocks in its last 700 with 0 rejects (13:47:09Z); the restart on its kept datadir at 13:47:15Z: the old process stopped at once (the new process's first lock line seven seconds after the marker; the watchdog held nothing, the b7cc37e7 fault closed), synced again at 13:48:39Z after 84 s, 109 templates read with max 3,432 ms and 0 timeouts; the kept read on pool-1's 0.3.17 copy on the same pod passed at 13:38Z (the rewrite line once, a clean second start). The pin's set on c4459193: the digest gate PASS, the mixed-version gate PASS, the wipe canary PASS, the kept read PASS, the restart PASS, the 12 GB line proves and verifies (paid is a race, not a gate); CASES END from c20-1 (about 14:50Z) is the last pin line. THE INTEROP FACT stands from the void run: the 5899f603 hub accepted 235 object-byte-5 blocks from the 8097d600 node with 0 rejected, one digest on all five nodes on the live sixteen-field file. The gates: the digest test and the kaspa-pow vector test (the amended devnet epoch-0 id 1a4230699a6b9c60 must equal, c120d7963abdcd96 must differ, the v3 control unchanged) on the box; the mixed-version Devnet 2 gate (the amended 0.3.20 node beside a 5899f603 node for ten minutes on the live file without the v4 fields) after the Mac build; the fresh-join canary the 0.3.20 cut's |
 | Main's rulings (7 October, morning) | no generator change to v4 on the live devnet; the record's null is the window model with numbers, sent by the hash lane to the attack-pass lane so AP-F8-1 re-gates against it; a fault beyond the model (a low-entropy source at site 15) stops at the coordinator with the two options priced (a 0.3.19 class amendment before the flip, or the flip held at the floor), nothing shipping without the founder's word; the tighter tail, an acceptance bound on the hot-set share, is a CLASS V5 item (sent to the v5 lane a6410f3b8abefb762 with the 64-seed census as its gate; the bound's number follows from the model) |