chore(pm): retire half-state row H25; H29 reports the awaiting-state pairs (#29626)

<!-- claude-projects-attribution: {"github_login":"huangyiirene"} -->
_Requested by **Irene** · [project
thread](https://claude.ai/code/project/chan_01TJXtQ1tVD26ecLDZxFS67x?thread=cmsg_01TJXtQ1tVD26ecLDZxFS67xCjGpvHD4tRyZ72DtPifmqL)_

Part of #29471

Clause-②: no
Metadata-protocol: none

## 维护者速读

- 改了什么:删除半态巡检行 H25(`pm:awaiting-maintainer` 与另一个 pm 状态标签共存),并删掉 H29 里「含
awaiting 的标签对归 H25」那一行跳过,让 H29 报出 H25 原来报的全部标签对。
- 为什么改:#29471 普查把 H25 列为一档(无读者、30 天无引用);席位裁决选项 C(评论 6589866189):只删
H25,H24 保留。
- 风险与回滚:只动 `scripts/pm/check-half-states.mjs`;没有哪个状态组合失去报告。回滚即 revert 本
PR。
- 先看:`h29PmStateExclusivity` 删掉的那行 `continue`,以及 H29 自测里由「静默」翻成「报出」的七条。
- 两分钟验证:`node scripts/pm/check-half-states.mjs --self-test`,预期 5202
条全过(主干 5219)。
- 需要你做的:无,席位落地。

## What changes

- **H25 is deleted**: its rationale block,
`AWAITING_MAINTAINER_CONFLICT_REASON` (only H25 read it),
`h25AwaitingMaintainerExclusivity`, its line in the header docblock, its
band-registry entry, its call in the sweep loop and the `h25row` test
helper.
- **H29 now owns the awaiting pairs.** The line `if (a ===
AWAITING_MAINTAINER_LABEL || b === AWAITING_MAINTAINER_LABEL) continue;
// H25's` is deleted, so H29 reports every pair that contains
`pm:awaiting-maintainer`. This is the one behaviour change to a staying
row, ruled by the seat (option C, [#29471 comment
6589866189](#29471 (comment))).
H29's H3 exclusion (`pm:queue` + `pm:dispatched`) is unchanged.
- **Kept**: `AWAITING_MAINTAINER_EXCLUSIVE_LABELS`, because
`PM_EXCLUSIVE_STATE_LABELS` spreads it; the `queued` fixture, which H3,
H8, H19, H22 and H29 cases use; H24 and the H8 case that names it, per
the ruling.
- **Prose pointers** that named H25 as a live row now name H29: the
`AWAITING_MAINTAINER_LABEL` reader list, the
`GOVERNED_PR_DECISION_LABEL` note, H55's header, the H66/H67 population
notes, the patrol input-set note, the sweep-loop comment and the H66
battery case's name. These are comment and test-name text only, with no
output change.
- No gate, row, constant or battery is added. No battery shrank, so no
pin moved.

## Why

The #29471 census put H25 in tier 1: no named reader and no 30-day
citation. The census grouped it with H24 as "the label pairs H29 already
covers", but on `main` H29 skipped every awaiting pair as H25's, and H24
is a label-and-assignee intersection that H29 never reads. The seat
narrowed the family to H25 and ruled that H29 takes over H25's pairs, so
no state loses its report.

### No state loses its report

Measured by importing both trees (`origin/main` 0a4e74e and this head)
and judging every non-empty subset of the six-label ONE-OF vocabulary as
an open card:

- 63 label sets judged. H25 fired on 31 of them, and the new H29 fires
on all 31 (0 misses).
- On all 63 sets, the answer to "does H3, H25 or H29 fire?" is the same
before and after.
- On every set without `pm:awaiting-maintainer`, H29's sentence is
byte-identical before and after.

## Verification

- `node scripts/pm/check-half-states.mjs --self-test` on 608b908:
**5202 cases pass** (5219 on `main`, so 17 fewer). The 17 removed are 14
H25 cases (9 single cases and the 5-case per-label loop), `H29: …and H25
does fire on it`, the H55 adjacency case that called H25, and the H55
reconciliation case that read H25's header prose.
- **The seven flipped cases**, each from "silent here, H25's" to "fires
here": `H29: a pair containing the awaiting state is this row's`, the
5-case loop `H29: awaiting + <state> fires here`, and `H29: …and names
the awaiting pairs too`.
- Every battery reads the same `n/floor` as on `main` (H66 182/172,
H70/H71 31/30 and so on), so nothing was re-pinned.
- Net lines: **+35 / −156 (−121)**, one file.

<details><summary>Local gates on 608b908 (`dispatch-gates
--commands`)</summary>

- On head 608b908, `node scripts/pm/dispatch-gates.mjs --commands`
derived 47 commands. All 47 ran, each with a 900 s timeout, with exit
codes recorded before any pipe. **All 47 exited 0.**
- The `--ran` reconciliation printed: `47 derived famil(ies) accounted
for — 47 run, 0 NOT-MEASURED`.
- `node scripts/check-commit-card-trailers.mjs --range
origin/main..HEAD`: exit 0.
- Prerequisite: `pnpm install --frozen-lockfile`.
- Families CI runs beyond the derived list are read from this PR's check
runs, not re-run locally.

</details>

## Acceptance notes

- `scripts/pm/ensure-pm-labels.sh:212` still names H25 in a comment as a
consumer of the label. That file was outside this card's surface, so it
was left alone. A one-word follow-up (H25 → H29) belongs to whichever
family next touches that file.
- H24 stays. It is the only row that reports an open `pm:queue` card
with an assignee, and H49, H50, H66 and H71 each decline that shape as
H24's.

## Human Minimum

- [x] 1. Schemas start as Zod; `packages/spec` holds no runtime logic.
(not touched)
- [x] 2. Every name I added uses the casing for its kind. (no names
added)
- [ ] 3. No formatter ran; `pnpm lint` passes. (no formatter ran; `pnpm
lint` was not run locally, so the CI lint job decides it)
- [x] 4. One issue, one branch off an up-to-date `main`, one pull
request.
- [x] 5. A change that publishes carries a changeset, and both
declaration lines are filled in here and in the changeset. (publishes
nothing: `scripts/pm/**` only; `skip-changeset`)
- [x] 6. `packages/spec` touched: its generated artifacts are
regenerated and committed. (not touched)
- [x] 7. No hand edits to `content/docs/references/`,
`packages/*/CHANGELOG.md` or `content/docs/releases/`.
- [x] 8. Title, body and comments are in English. (except the 维护者速读
block, which the seat contract writes in Chinese)
- [x] The *Before Submitting* list in CONTRIBUTING.md is done.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

https://claude.ai/code/session_01J8VVAmM3BRQQcQSA67VE3L

Co-authored-by: Claude <noreply@anthropic.com>

Author: objectstack-fleet[bot]