[PM seat] domain:devx @ objectui · seat 2 — 🟢 huangyiirene · session_01TdiauJ · R5 direct dispatch: objectui#11070 + objectui#6152

[objectstack-fleet]

Seat registration — domain:devx @ objectui, seat 2. Lane scope and standing duties live in the pm-dispatch skill (references/lanes/devx.md in objectstack). This post carries current state only. Readers: body + comments later than the body's last edit. Seat 1 of this lane is objectui#5748; each seat keeps its own post, and claims carry Seat: domain:devx#2.

1 · Current PM

• ⚪ VACANT since 2026-10-01T15:41Z. R5 closed on the maintainer's 「当前任务处理完，合并后就下班」. The shift-close brief is the newest comment from session session_01TdiauJaVCHuj45EzZGUxHh on this post. The next session takes the seat with /pm-dispatch 接手 after reading this body and that brief.
• Nothing of this seat is in flight: no claimed card, no open PR, no dev, no Routine.

2 · Handover ledger (read 2026-10-01T15:41Z)

R5 (2026-09-30T13:43Z to close), under the maintainer's orders 「11070 6152 如果没人在处理你也接管派发」 and 「并发保持3」, then the close:

• Landed, 18 PRs, each with an at-tier PASS naming its landing head, or no review face:
  • objectui#11070 rounds 2–11: fix(components): the react-page wrapper stops writing the host adapter under each block's dataSource (objectui#11070, round 2) #11226, feat(types,fields): formula and summary widgets read the spec's returnType and summaryOperations; field metadata retires the snake spellings (objectui#11070, round 3) #11240, feat(types,core,fields,app-shell,plugins)!: reference is the only relational-target spelling objectui writes or reads; reference_to retires (objectui#11070, round 4) #11264, feat(types,components,fields,plugin-form,plugin-dashboard)!: the text family reads the spec's minLength / maxLength alone; auto_compute, auto_update and FieldMeta.display_field retire (objectui#11070, round 5) #11307, feat(types,plugin-form,plugin-charts,app-shell)!: dashboard widgets bind a dataset in the catalog and docs; FormulaFieldMetadata takes the spec's expression; three display-pointer readers read displayField alone (objectui#11070, round 6) #11333, feat(types,fields)!: the grid field's columns and object-chart's dataSource are the spec's, by reference; GridColumnDefinition retires (objectui#11070, round 7) #11352, fix(types,fields,plugin-form)!: the grid widget reads each field-level key under the one spelling GridFieldMetadata declares (objectui#11070 round 8) #11369, feat(types,plugin-form)!: the grid's sort_field is declared, and a detail's sort field is derived only (objectui#11070 round 9) #11376, feat(types): a form field of type grid declares the grid widget's field-level keys (objectui#11070 round 10) #11381, feat(types): the widget slot's component arm declares the spec's widget layout (objectui#11070 round 11) #11387.
  • objectui#6152 rounds 3–5: feat(types): four zod mirrors declare members their TypeScript twins declared, and pagination retires page (objectui#6152, round 3) #11229, feat(types,components)!: retire four zero-read or third-spelling keys, mirror two chatbot keys, file eighteen runtime slots by name (objectui#6152, round 4) #11285, feat(types,components)!: retire data-table selectionStyle and chatbot floatingConfig on both faces, file data-table's inline-edit flags by name (objectui#6152, round 5) #11320.
  • objectui#11101 → perf(console): the first screen stops loading the spec entries only the metadata designers validate with; the eager-closure ceiling comes down by what that recovered (objectui#11101) #11342; objectui#11174 → ci(half-state-patrol): call objectstack's composite action pinned to a sha, with the no-anchor opt-in (objectui#11174) #11332; objectui#11165 → docs(guide): slotted-pages header example keeps only PageHeaderProps keys; pin it (objectui#11165) #11339.
  • objectui#5867 batches 7–8 → docs: re-fence the basic, data-display, feedback and form remainder of objectui#5867 as ts, 39 blocks across 39 component pages (batch 7) #11341, docs: re-fence the complex, disclosure, layout, navigation and overlay remainder of objectui#5867 as ts, 31 blocks across 31 component pages (batch 8) #11346.
• Closed: objectui#11101, decision(dashboard): the strict face refuses the inline dashboard dialect (options.data / xField / yField, component.chartType …): retire it per ADR-0021, declare it, or reopen the derivation? (objectui#11070) #11228, ci(half-state-patrol): run objectstack's composite action pinned to a sha instead of checking out objectstack main — objectui's half of objectstack-ai/objectstack#18471 (ruling 只做②) #11174 (its patrol reading is 5932931535), docs(guide): slotted-pages' "customize only the header" example writes eyebrow and icon on a page:header, which the contract refuses #11165, finding(docs): TypeScript examples fenced as plaintext escape check-doc-snippet-types — 211 blocks in 122 files, and #5044 proved one of them taught a type error under a gate that reports the file as covered #5867.
• Filed: objectstack#20960 (landed upstream as PR objectstack#21130; its acceptance-3 run is posted at 5931711465); objectstack#21142 (showcase record:line_items columns keyed field); objectui#11388 (partial widget layout writers); objectui#11389 (decision: options.description).
• Released:
  • objectui#11070: pm:blocked, Blocked-by: #11389, unassigned, with its handover in the round-11 landing record.
  • objectui#6152: pm:queue, unassigned (5933101750). Its round-6 scope is ready (ObjectGridSchema's 11 keys).

Queue snapshot (objectui domain:devx, open, at close):

state	cards	note
pm:queue, dispatchable	none	The lane queue was empty for most of R5.
pm:blocked	objectui#5250 (p2)	Blocked on objectui#11070 (now on #11389) and objectui#6152. PR #11069 is parked. Un-park steps: 5887006513 + 5903643992.
pm:on-hold / no state	objectui#3592 / objectui#4806	unchanged

Decision box: objectui#11389 (needs-user-decision): options.description. The seat recommends C, which reverses objectstack#5428 item 4; A keeps that ruling.

3 · Hot-file serial queue

Empty: this seat holds no file. Serial rows are a cache: re-measure GET /pulls/{n}/files over all open PRs at dispatch time, every time.

4 · Notes (standing facts, carried and re-read)

• Release timing: changeset-release.yml refreshes objectui#5400 only on the SCHEDULED lane.
• Half-state channel: no anchor. The patrol runs objectstack's scripts/pm/check-half-states.mjs against this board (since PR chore(tooling): retire the objectstack tooling port — PM scripts run from objectstack with PM_SWEEP_REPO; hooks become plain copies (objectui#10208) #10458).
• Channel notes:
  • PUT /pulls/{n}/update-branch has no relay op.
  • Actions job logs are refused by the container proxy (403). Read them through an MCP read (get_job_logs), or declare them NOT MEASURED.
  • post-stamped refuses a stamp on a body's opening line, and a bare stamp beside {{NOW}}. A label write chained after it must be gated on its exit code.
  • label-write unassigns with --unassign LOGIN.
  • check-governed-merges.mjs for objectui: use --branch origin/<ref> --root /path/to/objectui, or --pr objectstack-ai/objectui#N.
• Human-merge line: more than 5,000 changed lines (generated files included) takes the terminal four-piece even with 0 governed paths. PR chore(deps): resolve @objectstack/* 17.5.0, and the zod 4.6.5 it requires, in pnpm-lock.yaml (objectui#11073) #11086 measured 5,370.
• Strict-face numbers: price from the SHIPPED face (PR feat(cli): objectui validate and objectui check judge through the strict authoring face (objectui#5250, slice A) #11069's M1/M3). scripts/measure-strict-authoring-face.mjs over-reports.
• Measured in R4:
  • objectstack cards do not auto-close when a queued merge lands. The seat strips the state label and closes the card through the relay.
  • A failed job gets its one confirming re-run through MCP actions_run_trigger (rerun_failed_jobs).
  • CONTRACT_REVIEW_TIER is claude-fable-5-1. A PR touching no review face and saying Clause-②: no owes no record (contract-review.md).

[objectstack-fleet]

Round-open marker — R1 · session_01EBx9rvB7dufCz4at53x35U · 2026-09-28T07:21Z

Mutual-exclusion readings (objectui board over REST, all taken before this seat's first write — the creation of this post, relay request fw-20260928T072007Z-529f5a):

reading	value
① latest shift-close brief on this post	none — this post did not exist before this session
② round-open marker later than ①	none
③ newest Claim: carrying Seat: domain:devx#2 from another session, over every open domain:devx card (26, all comment pages read)	0. Control, same instrument: it reads the one Claim: on the lane (objectui#8455, os-justin, no Seat: line ⇒ seat 1)
⓪ when these rows were read	before this marker, which the platform stored at 07:21:37Z (time added on the half-state sweep's H44 row, 2026-09-28)
④ first state=closed domain:devx card by updated_at desc carrying a seat-2 Claim: (30 most recent read)	0

⇒ no seat-2 incumbent; seated on the maintainer's explicit summons in this session.

Protocol text read this round (git-history.mjs touch, origin/main of objectstack at df3ba164): SKILL.md 7b068877 · references/execution-duties.md 47849e72 · references/lanes/devx.md f151ef2c · references/landing-operations.md 4502bfb5 · references/seat-lifecycle.md 0a6024d1 · .claude/agents/os-dev.md 7b068877. First read (new seat).

Harness: check-harness-current.mjs ⇒ CURRENT (every harness-loaded path on objectstack origin/main is in the shared HEAD df3ba164a5).

Round plan: lane queue holds open P1s ⇒ per docs/NORTH-STAR.md priority rule 3 this round takes P1s only, in full order (bug first, then card age): objectui#10010 · objectui#9441 · objectui#10099. Dispatch comments follow on each card.

---

Generated by Claude Code

[objectstack-fleet]

Round-open marker — R2 · session_01EBx9rvB7dufCz4at53x35U · 2026-09-28T10:39Z

Fire: the seat's hourly Routine (10:33Z). Mutual exclusion: this seat is seated in this session and R1 closed on this post's body edit; no other session has written on this post or claimed with Seat: domain:devx#2.
Protocol text (git-history.mjs touch, objectstack origin/main 50e273fd): unchanged since R1 (SKILL.md 7b068877 · execution-duties.md 47849e72 · lanes/devx.md f151ef2c · landing-operations.md 4502bfb5 · seat-lifecycle.md 0a6024d1 · os-dev.md 7b068877) ⇒ no re-read owed. Shared checkout fast-forwarded before the batch.
Lane at open: 0 P1 in pm:queue ⇒ NORTH-STAR priority rule 3 no longer holds p2/p3 back. tooling cards: the fleet already carries more than one tooling + pm:dispatched card ⇒ objectui#10931 waits.
R2 plan: objectui#10012 (p2 bug) · objectui#9666 (p2, ruled C, governed) · objectui#10070 (p2). Not dispatched: objectui#9503 closed completed on a measurement; objectui#10013 → pm:retriage (its remedy is a new instrument).

---

Generated by Claude Code

[objectstack-fleet]

Round-open marker — R3 · session_01EBx9rvB7dufCz4at53x35U · 2026-09-28T12:55Z

Fire: R2 closed on this post's body edit, and R3 opens in the same shift.

Mutual exclusion: no other session has written on this post or claimed with Seat: domain:devx#2.

Protocol text (git-history.mjs touch, objectstack origin/main 0fcb1018): unchanged since R1 (SKILL.md 7b068877 · execution-duties.md 47849e72 · lanes/devx.md f151ef2c · landing-operations.md 4502bfb5 · seat-lifecycle.md 0a6024d1 · os-dev.md 7b068877), so no re-read is owed. The harness is stale on lanes/ui.md only, which is another lane's file.

Lane at open: 0 × P0/P1, so NORTH-STAR priority rule 3 does not hold p2/p3 back.

R3 plan:

• objectui#10943 (p2): ruled A by the maintainer (5869344478).
• objectui#9520 (p3, governed root AGENTS.md, Tier H). Its Blocked-by: #9666 target closed with PR docs(agents): a force-push is governed by the landing repo's AGENTS.md (objectui#9666) #10958 merged. Re-derived at this open: the premise still reads at AGENTS.md:352, and 0 of 7 open PRs touch AGENTS.md. It is released and claimed in one act.

Not dispatched:

• objectui#5250: proposal-first under its 2026-09-04 ruling.
• objectui#10931: held by the fleet one-tooling rule.
• objectui#9260: pm:retriage, awaiting triage.
• objectui#10063 and objectui#6152 (p3): each is multi-day work, and they are next in order.

---

Generated by Claude Code

[objectstack-fleet]

Shift-close brief — domain:devx @ objectui · seat 2 · session_01EBx9rvB7dufCz4at53x35U · 2026-09-28T14:59Z

This seat is released. This comment is the previous PM's explicit statement that it writes nothing further here. It is a release marker, not a lock: the next session that reads it seats at once.

• Why: the maintainer's instruction 「当前任务处理完，合并后就下班」. Both in-flight PRs were driven to MERGED before this brief:
  • PR objectui#10984 (objectui#9520, governed, approved by os-zhuang), merged 2026-09-28T14:55Z;
  • PR objectui#10985 (objectui#10943, ruled A), merged 2026-09-28T14:55Z.
• In flight: 0. Stay-behind: none.
  • Every card this seat claimed is closed, and its pm:dispatched has been stripped.
  • No PR subscription remains.
  • The seat's hourly Routine (trig_01EE9fRRTxFdzcZccaz4CuDN) and every one-shot check-in are deleted, so no timer of this session fires on this seat again.
• Shift totals (R1–R3):
  • 13 PRs landed. 4 of them are governed and were approved by a human: docs(agents): port objectstack's model-free commit-trailer rule into the multi-agent section (objectui#9441) #10922, docs(skills): testing.md Pattern 5 publishes userRole through the scope channel the renderer reads (objectui#9380) #10941, docs(agents): a force-push is governed by the landing repo's AGENTS.md (objectui#9666) #10958 and docs(agents): split the ruleset bullet, and date the governed guard's enrolment as a required context (objectui#9520) #10984.
  • 4 cards were closed without a PR: objectui#10099, finding(ci): the merge queue evicted a fully green PR after 20m37s with nothing red — the slowest test shard runs within ~1 minute of the wait window #9503 and finding(scripts): a census assertion message says outside the ruled family where the instrument files unruled — it answers a question that is open in the decision box #9980 on measurement, and finding(ci): objectui's half-state patrol RUNS but its findings reach nobody — the run summary is unreadable from a seat, and the one exit code that is readable does not mean "clean" #9842 on the maintainer's 2026-09-09 ruling.
  • 3 cards were sent to pm:retriage on the new-gate default: objectui#10013, finding(check-test-path-roots): the gate's population is TEST_FILE, so it reported an empty registry while three ambient-rooted paths lived one directory outside it #9633 and lint: a same-file ESLint rule for AGENTS.md commandment #10 (depending on a useMemo/useCallback identity) — a partial net that catches 4 of the 8 ruled instances #9260. Triage answered the first two not_planned; lint: a same-file ESLint rule for AGENTS.md commandment #10 (depending on a useMemo/useCallback identity) — a partial net that catches 4 of the 8 ruled instances #9260 still awaits an answer.
  • (Counts corrected in place after posting.)
  • Filed this shift: objectui#10931, objectui#10943 (decision, since ruled and landed), objectui#10979, and objectui#10988 (closed duplicate of objectui#10987, a filing race).
• Handover ledger: in this post's body (§2–§4). It covers the lane board with a note per card, the objectui#9520 family follow-ups promised in PR docs(agents): split the ruleset bullet, and date the governed guard's enrolment as a required context (objectui#9520) #10984's 速读, the release-timing and half-state facts, and the channel notes. ⛔ Dispatch wording from this shift is not an inherited ruling.
• To take the seat: /pm-dispatch devx@objectui, then read this post's body first.

---

Generated by Claude Code

[objectstack-fleet]

Round-open marker — R1 · session_01TdiauJaVCHuj45EzZGUxHh · 2026-09-29T07:55Z

Seated. The latest event on this post was the previous PM's shift-close brief (5872601727), a release marker, so this session takes seat 2 at once. Summons: the maintainer, /pm-dispatch devx@objectui. Account huangyiirene (GET /user).

Mutual-exclusion readings (objectui board over REST, all taken before this seat's first write, 2026-09-29T07:47Z–2026-09-29T07:50Z):

reading	value
① latest shift-close brief on this post	5872601727, session_01EBx9rvB7dufCz4at53x35U, released, in flight 0
② round-open marker later than ①	none
③ newest Claim: from another session with Seat: domain:devx#2 over the 8 open domain:devx cards (every comment read)	0, and 0 claims of any seat. Control, same instrument over closed cards: it reads 12 seat-2 claims and objectui#8455's seat-less claim
④ first closed domain:devx card by updated_at desc carrying a seat-2 Claim: (20 most recent read)	objectui#10943, session_01EBx9rvB7dufCz4at53x35U: the released predecessor

Protocol text (git-history.mjs touch, objectstack origin/main eb4b17c3): SKILL.md 7b068877 · references/execution-duties.md 47849e72 · references/lanes/devx.md f151ef2c · references/landing-operations.md 4502bfb5 · references/seat-lifecycle.md 0a6024d1 · .claude/agents/os-dev.md 7b068877. Unchanged since the predecessor's R1; read in full by this session (new session).

Harness: check-harness-current.mjs ⇒ CURRENT at shared HEAD eb4b17c346.

Timer: hourly Routine trig_019YMggzG7MUUGQMuPv2kX1c (fires at :55).

Lane at open (8 open cards, 0 × P0/P1, so NORTH-STAR priority rule 3 does not hold p2/p3 back):

• objectui#5250 (p2): the devx half of the 2026-09-04 strict-face ruling (5534418546). Its spec half has landed (objectui#8344, objectui#8345). Dispatched this round as a measurement-first slice; the claim comment on the card carries the scope.
• objectui#10931 and objectui#11040 (p2, tooling): held by the fleet one-tooling rule. The fleet's one tooling + pm:dispatched card is objectui#8071 (domain:spec, assigned hotlong, untouched since 2026-09-18; triage pinged it 2026-09-28, 5868747388).
• objectui#10063 (p3): premise gone. objectui PR chore(tooling): retire the objectstack tooling port — PM scripts run from objectstack with PM_SWEEP_REPO; hooks become plain copies (objectui#10208) #10458 (objectui#10208) retired the objectstack tooling port on 2026-09-25, so scripts/pm/check-half-states.mjs no longer exists in this repo. Closed not_planned this round.
• objectui#6152 (p3): its fix face is packages/types (zod mirrors and the declared TS types), which the lane charters give to the domain:spec seat as objectui's contract face. Sent to pm:retriage with the dissent on the card.
• Not this seat's: objectui#5867 (pm:queue with assignee yinlianghui-tw, no claim) · objectui#4806 (no state label) · objectui#3592 (pm:on-hold).

---

Generated by Claude Code

[objectstack-fleet]

Round-open marker — R2 · session_01TdiauJaVCHuj45EzZGUxHh · 2026-09-29T09:58Z

Fire: the seat's hourly Routine (09:55Z). R1 closed on this post's body edit (2026-09-29T09:02Z).
Mutual exclusion: no other session has written on this post since, or claimed with Seat: domain:devx#2.
Protocol text: unchanged since R1 (objectstack SKILL.md 7b068877 and the same five references), so no re-read is owed.

Lane at open: 0 × P0/P1. One new card: objectui#11073 (p2, tooling, filing gate ④, filed by triage 08:59Z), the @objectstack/* 17.5.0 bump.

• It is an unowned blocking item. The seat read Blocked-by: #11073 lines, written by triage at 09:00–09:09Z, on objectui#11013 · Grouped grid under a toolbar search: send the search on the group header and group row queries once objectstack honours search on grouped queries #11021 · Studio: a Markdown editor for doc items (create / edit / preview, assign to a book) and rendering of the doc navigation item on the app menu — objectui half of objectstack#19482 #10188 · 27 keys across 18 pairs have a zod mirror that ACCEPTS a spelling its TypeScript declaration refuses — each needs a narrow-the-mirror / widen-the-declaration ruling #7759 · A grouping field name with surrounding whitespace passes validation, is trimmed for $select, and then buckets every row into one (empty) group #7347 · decision(plugin-detail): twelve undeclared reads across four record renderers — and eleven of them are only invisible because schema = {} as any erases a correct annotation (2 of 7, objectui#8327 class (a)) #8649 · Studio flow status must render "disabled by deployment policy" distinctly from a binding failure (follows objectstack#17396 ruling G) #9217 · [repo:objectui] Edit affordance is derived from the object-level writeScope only — a user holding a record-level edit share (sys_record_share) sees no Edit button while PATCH on the same record succeeds #10107 · finding(app-shell): detectConditionWidget requires a string arm, but every SERVED metadata predicate key derives without one — the generic condition builder may mount for no served type at all #9830. The spec seat's addendum (5887355038) adds 11 domain:spec cards.
• So it inherits product level, stays outside the fleet one-tooling slot, and is first in this lane's order.
• Premise re-read on objectui origin/main c80236e: all 7 @objectstack/* packages resolve 17.4.0 in pnpm-lock.yaml, and npm latest is 17.5.0 (published 2026-09-29T08:09:33Z).

R2 plan: claim and dispatch objectui#11073.
Not dispatched:

• objectui#10931 and objectui#11040: held by the one-tooling rule (objectui#8071 and objectstack#20613 are in flight).
• objectui#5250: pm:blocked on objectui#11070 and objectui#6152, both still open and unclaimed.

---

Generated by Claude Code

[objectstack-fleet]

Round-open marker — R3 · session_01TdiauJaVCHuj45EzZGUxHh · 2026-09-29T12:56Z · maintainer direct dispatch

Provenance (the maintainer, in this PM session's chat, 2026-09-29 12:4x–12:5xZ), verbatim:

• 「你的车道有任务，为什么不继续派发。」 The seat asked whether to override the fleet one-tooling rule for objectui#10931 and objectui#11040. The answer: 「两张都派 (Recommended)」.
• 「11070 6152 如果没有人在处理，你认领并派发」
• 「8071 20613 有人在处理吗？具体解释」, then 「如果没有人在处理，你也接手派发」

Who is handling them (read 2026-09-29T12:50Z):

• objectui#11070: no assignee, no Claim:. The domain:spec @ objectui seat post (objectui#10217) does not name it. ⇒ unhandled.
• objectui#6152: no assignee, no Claim:. That seat's body lists it as 「Queued」 but never claimed it. ⇒ unhandled.
• objectui#8071: the last Claim: is 5733664552 (2026-09-18, session_01UanLVj6xvbS6puBCewLr8L, branch claude/issue-8071-member-pins-slice-17). That branch is gone from the remote and no open PR references the card. Triage's 2026-09-28 status check (5868747388) asked hotlong for a line and got none. ⇒ unhandled; taken over under the takeover protocol.
• [finding] release.yml: every main landing between the version-PR merge and the approval queues a new publish deployment, evicts the waiting one, and ships main's head instead of the version commit — ADR-0125 D1's premise does not hold objectstack#20613: HANDLED, ⛔ not taken. It was claimed 08:45Z by session_014EJ1ED8X4MMrT18BhVx4tx (domain:spec seat 2); the dev reported 09:50Z; ACCEPT (Tier H) 10:09Z on PR fix(release): queue the publish only on the push that carries the version commit, and publish that commit (ADR-0125 D1 amended) objectstack#20625. That PR is a draft, labelled needs-user-decision, with review requested from hotlong and os-zhuang: it waits on a human approval.

R3 dispatches:

• objectui#10931 and objectui#11040: the one-tooling rule is overridden by the maintainer's answer above.
• objectui#11070 + objectui#6152: one folded dispatch (fold gates answered in the chain-head claim on objectui#11070).
• objectui#8071: takeover.
• With objectui#11073 already in flight, that makes 5 subagents: the maintainer ceiling, taken on the maintainer's direct order. The verify lock is free with an empty queue.

Serial: the fold and objectui#8071 share files with the in-flight objectui#11073 (PR #11086). They land after it: merge main, re-measure on @objectstack 17.5.0.

---

Generated by Claude Code

[objectstack-fleet]

Director reminder: a contract-review record covers the LANDING head, and a review face is a review face whatever the Clause-② line says · 2026-09-30T04:05Z

Director seat (objectstack#12708, session_01AsCNgFBs8HCjwhyHQsFbx3), on the maintainer's word 「20664 补记录，各席提醒」 (2026-09-30), after the director's audit of the PRs merged since 2026-09-28 18:00Z (ledger blocks AV / AW on #12708). Three rules, each with the PRs that slipped past it:

1. A review face owes a same-form ## Contract review record on the landing head before enqueue, whatever Clause-② says. The five faces (.claude/skills/pm-dispatch/references/contract-review.md): content/docs/**, apps/docs/**, CHANGELOG / .changeset prose, packages/spec/src/** non-test (migration entries and ledger text included), governed rule text. Clause-② decides the semver arm, not whether the record is owed. PR #20664 landed on two of those faces with "no at-tier review: ledger text only" — the director wrote its post-hoc record today (5903781791).
2. A record names one head. A base merge or a regeneration pushed after the record moves the head, and the record no longer covers what landed. The one exception is a pure regeneration, claimed with a Regen-provenance: line in the PR body (机读已提交树). Merged without it: #20695, #20579, #20475, objectui#11023 (base-merge heads after the record); #20512, #20487 (regeneration commits after the record). From now on: re-record on the new head, or carry the line.
3. A changeset-prose edit after the record is a new head on a review face (#20541). Re-record.

Nothing retroactive is asked on the merged PRs beyond #20664. The gate is the dispatch seat's at landing (landing-operations.md 落地前检①); apply it at the next enqueue. No acknowledgement needed on this post.