diff --git a/docs/remediation/BOARD.md b/docs/remediation/BOARD.md index d276a425..9ea0aa1f 100644 --- a/docs/remediation/BOARD.md +++ b/docs/remediation/BOARD.md @@ -1,7 +1,9 @@ # mos-remediation — LIVE BOARD (keep < 8 KB) -**Phase:** EXECUTING — RM-03 at owner-merge; RM-61 **MERGED**; **RM-02 keystone is the front**. -**Updated:** 2026-08-01 — seam crossed; successor seat resumed, attested from the files, and is driving. +**Phase:** EXECUTING — **RM-02 keystone is the front**. +**Updated:** 2026-08-05 — successor seat live. **State correction (D-56):** round-4 verdict id 73 had +landed 2026-08-01T18:12Z (NO-GO @ `fbb61912`); remediation was pushed the same evening; the round-5 +re-review dispatch was lost at the seam and the lane idled 4 days. Round 5 re-dispatched 2026-08-05. ⚠ That was a **MANUAL pane respawn** (prior seat ~803k tokens): it validates the checkpoint+rehydration **design**, NOT a lifecycle **mechanism** — P-LIFECYCLE rotation does not exist yet (**D-41 / RM-62**). @@ -16,14 +18,14 @@ ## In-flight -| Task | Owner | State | -| ------------------- | --------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | -| RM-01 checkout | — | **MERGED** `f58b3699` (#1027) | -| RM-03 queue guard | Jason (re-sync) | ✅ **MERGED** `58b971ab` (#1032), #1019 closed. ⚠ **NOT DELIVERED**: installed guard still the broken one (291 lines / 0 `ASSERTED_NOT_READY` vs main 482 / 5). Re-sync via `mosaic upgrade`, then **prove it blocks a KNOWN-RED pipeline** — **D-51** | -| RM-02 registry ★key | rev-974 | **RE-REVIEW @ `e910a45a`** — narrowing round; overclaim control fires @84 (verified). ⚠ **3 branch authors — squash erases 2 (D-55).** Trade + trailers recorded in body; **branch MUST NOT be deleted**. ACs @ `dde38717` | -| RM-61 CI exemption | — | ✅ **MERGED** `f4fd5967` (#1033). #1034 closed; **#1000 stays OPEN** (retirement trigger). Exemption is on `main` | -| RM-59 / RM-60 | Jason (infra) | tracked deps; RM-60 option **B** | -| #1023 queue attempt | Jason | SUPERSEDED-PENDING-JASON — live REQUEST_CHANGES, do **not** merge | +| Task | Owner | State | +| ------------------- | --------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | +| RM-01 checkout | — | **MERGED** `f58b3699` (#1027) | +| RM-03 queue guard | Jason (re-sync) | ✅ **MERGED** `58b971ab` (#1032), #1019 closed. ⚠ **NOT DELIVERED**: installed guard still the broken one (291 lines / 0 `ASSERTED_NOT_READY` vs main 482 / 5). Re-sync via `mosaic upgrade`, then **prove it blocks a KNOWN-RED pipeline** — **D-51** | +| RM-02 registry ★key | rev-974 | **ROUND-5 RE-REVIEW DISPATCHED @ `e910a45a` (2026-08-05, PR comment + verified tmux — D-56).** Round-4 id 73 NO-GO @ `fbb61912` live/undismissed; remediation `9b7005d5..e910a45a` pushed 08-01. ⚠ **3 branch authors — squash erases 2 (D-55);** explicit squash + trailers (pr-merge.sh field: be-coder-08); trade in body; **branch MUST NOT be deleted**. ACs @ `dde38717` | +| RM-61 CI exemption | — | ✅ **MERGED** `f4fd5967` (#1033). #1034 closed; **#1000 stays OPEN** (retirement trigger). Exemption is on `main` | +| RM-59 / RM-60 | Jason (infra) | tracked deps; RM-60 option **B** | +| #1023 queue attempt | Jason | SUPERSEDED-PENDING-JASON — live REQUEST_CHANGES, do **not** merge | ### For the incoming orchestrator — read this before acting diff --git a/docs/remediation/TASKS.md b/docs/remediation/TASKS.md index af7c89c4..e23b8e41 100644 --- a/docs/remediation/TASKS.md +++ b/docs/remediation/TASKS.md @@ -586,6 +586,48 @@ is that discipline without a mechanism eventually fails."_ > a failed round **or by a lucky manual sweep.** Interim: both coordinator and orchestrator sweep — and > **the mechanism retires the sweep.** +### D-56 — the lane idled FOUR DAYS on a verdict that had already landed: a "waiting on X" state line decays like any other measurement + +**Caught by Mos (2026-08-05), provider-measured by this seat the same tick. The failure is shared: +the board carried it, the continuity file carried it, and the coordinator relayed it all night.** + +Facts, all measured at the provider on 2026-08-05: + +- Round-4 verdict **review id 73** — `rev-974`, REQUEST_CHANGES, bound `fbb61912`, submitted + **2026-08-01T18:12:03Z**, live/undismissed. Three blockers: A1 local-ref anchor, P2 source-anchor + shrink-together, D-38 evidence-subject tautology. Four earlier rounds dismissed. +- `coder-mos2` pushed remediation commits `9b7005d5` → `c5b0d510` → `e910a45a` by ~20:10Z **the same + evening**. Each blocker has a corresponding change (history anchor excised and held at the RM-60 + boundary per `GATE-CLAIMS.md`; `gate-inventory-shrink-control` gains a `source-manifest` + shrink-together mode; `gate-population-control` gains an `evidence-subject`-at-consumption mode). + Whether they SUFFICE is the reviewer's ruling, not this file's claim. +- **No round-5 re-review request is observable anywhere.** The only PR comment since the verdict is + the verdict itself; `rev-974` produced no review at `e910a45a` in four days and was found alive, + tasked on an unrelated lane. The board's "RE-REVIEW @ e910a45a" recorded an intent whose dispatch + was never confirmed delivered — the 2026-08-01 rotation seam is the likely loss point. +- Net: the lane sat "blocked on rev-974's verdict" while the verdict it was actually waiting on had + **never been requested**, and the verdict everyone could see had **already landed**. + +**The general form (Mos's, owned by him, kept verbatim in substance):** a "waiting on X" state line is +a **claim about X's behaviour** and decays like any other measurement. Heads, verdicts and CI are +re-verified at the moment of action; "blocked on" is carried forward across rotations without ever +being re-read — **and a rotation is exactly when it must be re-measured, because the successor +inherits the belief with no way to date it.** + +Two requirements fall out: + +1. **A dispatch is not "sent" until receipt is observed** — the charter's message-delivery clause, + applied to review dispatches specifically. `agent-send` "queued" is not "processed". Every review + dispatch MUST leave a **durable provider-side record** (PR comment) in addition to the tmux + message, so a lost send is detectable by anyone from the artifact rather than from memory. + Binding on this seat's process now; on the coordinator daemon (RM-58/RM-63 class) when it exists. +2. **Rehydration briefs must mark every inherited "blocked on X" as PROVENANCE, not fact** — the + continuity file already says this for SHAs; it now explicitly covers waiting-state lines, which + are the same class one level up. + +**Fix applied same tick:** round-5 re-review dispatched to `rev-974` at `e910a45a` — durable PR +comment plus verified tmux delivery — and the board/continuity corrected. + ### D-52 — the FIFTH arrival defeated the orchestrator AND the coordinator, in the same turn, on a principle they had both just written `coder-mos2` implemented blocker 2's fix — a `gates/required-gates.baseline.json` "protected/base