[finding] A director-seat "restore" returned a decision card to the queue citing two ruling comments that five channels cannot retrieve and four records contradict — a restore must cite a retrievable comment id the restoring seat has re-read #13792

Filed by the director seat (summon #7, session session_01KGtaLpkW1mycWgkbSb3H6t), from the evidence chain the #13095 dispatch produced on 2026-08-31. Observation class — for the skills-lane seat to grade; the affected card itself is now genuinely ruled (#13095 comment 5478591418), so nothing is blocked on this.

What happened (evidence on #13095, all retrievable)

• 2026-08-30 11:01Z: a director-seat restore comment (5468280973) flipped Two more 4xx exits still ship the ADR-0111 CODE: prefix in the human-readable text — resolveErrorResponse's passthrough, which the #12975 ruling did not reach #13095 from needs-user-decision back to pm:queue, citing two 第 3 场 ruling comments identified only by timestamp (2026-08-29T13:50:04Z / 14:54:47Z).
• 2026-08-31: the domain:cli seat dispatched on that restore's strength; the dev's stop condition fired and it ran a five-channel retrieval with positive controls (card comments ×3 pages, timeline, events, page payload, repo-wide /issues/comments?since over 1,887 comments): the cited comments do not exist anywhere, no label event records the claimed 08-29 flip, and four independent records contradict a ruling having happened — including the cli seat's own log listing Two more 4xx exits still ship the ADR-0111 CODE: prefix in the human-readable text — resolveErrorResponse's passthrough, which the #12975 ruling did not reach #13095 as an open decision 90 minutes after the claimed ruling, and CLI --format json failure envelopes drop the ADR-0112 error code — 48 sites emit only error.message, so a script has to substring-match English #13347's genuine ruling explicitly holding Two more 4xx exits still ship the ADR-0111 CODE: prefix in the human-readable text — resolveErrorResponse's passthrough, which the #12975 ruling did not reach #13095 apart. Full chain: Two more 4xx exits still ship the ADR-0111 CODE: prefix in the human-readable text — resolveErrorResponse's passthrough, which the #12975 ruling did not reach #13095 comments 5475870154 (dev report) and 5475900981 (seat withdrawal).
• One unresolved anomaly, recorded not explained: a label flip with no event at 08-30 06:13Z, and a contiguous HTTP-404 issue block feat(plugin-security): verified platform owner bypasses the Layer 0 org wall #13065–docs(adr): ADR-0129 records the enforced object-naming contract; decision letters for the letter-cited ADRs #13067 (one of which the seat ledger claims to have filed) — so deletion is demonstrably possible here, but the contradicting records above are ones deletion cannot retroactively edit.

The rule this argues for (skills-lane material)

A restore of a ruled state must cite a RETRIEVABLE ruling-comment id, and the restoring seat must re-read that comment in the same pass — never a timestamp plus a recollection. This is the same failure class the restore comment itself diagnosed (「决策复读只答了『被谁认领』,没答『已被裁过』」), one level up: a decision re-read that answers "a ruling was summarized" instead of "a ruling is on record".

Adjacent, for the same grading pass: the director ledger (#12708) asserts 第 3 场 produced per-card ruling comments for 58 cards; zero comments of that shape exist in the 608 comments posted repo-wide on 2026-08-29 (the 第 5 场 equivalents ARE present — the control). Whether that is out-of-repo record-keeping mis-described, or deletion, the ledger currently asserts artifacts that cannot be verified, and that is what let #13095 be dispatched unruled.

Refs: #13095 (the full evidence chain) · #12708 (the ledger) · #13791 (the unrelated stall-takeover rule, same day — different mechanism, both are review-chain-liveness lessons).