docs(queen): a redispatch after a spent ceiling gets one attempt, not one more set

[dmitrii-f-t27]

Closes #1769

What

Option (b) of #1769: the behaviour stays, and two comments now state what the code does.

• queen-tick.ts, above MAX_CEILING_RELEASES: it said the bee gets "one more set" of attempts after a spent ceiling. It now says one more attempt, and explains why.
• queen-dispatch.ts, the redispatch upsert: it said the new attempt "is judged on its own work". It now says that send_backs restarts at 0, but ceiling_releases is already at MAX_RELEASES, so the first send-back of the hand-back is closed as obsolete by the review valve.

Mechanism (unchanged, on queen @ fc0b59d): the redispatch sets ceiling_releases = 1. The next send-back goes to nextStep(KIND_SEND_BACK_CEILING, idleMinutes, releases). next_step in specs/queen/review_valve.t27 returns STEP_CLOSE once releases >= MAX_RELEASES (MAX_RELEASES = 1). The claim becomes cancelled.

Scope

• Comments only. No code, SQL, spec or test changes. Both files are already on queen.
• No .t27 spec change: no behaviour changes. The behaviour is already pinned by specs/queen/claim.t27 in spec(queen): claim.t27 - the round's stateOfDispatch as a card, the review valve imported (Closes #8765) t27#8768 (test a_send_back_under_its_ceiling_goes_back_after_the_hour).
• Option (a), the behaviour change, is tracked in Queen: a redispatch after a spent ceiling should get a full set of send-backs (option (a) of #1769, depends on t27#8768) #1806. It depends on spec(queen): claim.t27 - the round's stateOfDispatch as a card, the review valve imported (Closes #8765) t27#8768 and feat(queen): the round's claim is decided by the claim card, proven equal to the hand-written rule first (Closes #1767) #1768.
• git merge-tree against feat(queen): the round's claim is decided by the claim card, proven equal to the hand-written rule first (Closes #1767) #1768 (head 9a9e124): no conflict.

Checks

• Both edited files parse with Bun 1.3.6 (bun build --no-bundle). The SQL comment stays inside the template literal (no backticks added).
• CI on this PR: see the checks below.

Dmitrii Fedorov (@dmitrii-f-t27)

[tri-claw]

Summary by t27-bees

This PR updates comments in two files to clarify that a redispatch after a spent ceiling grants exactly one more attempt (not "one more set" of attempts). The comments now explain the mechanism: send_backs resets to 0 but ceiling_releases is already at MAX_RELEASES (1), so the first send-back hits the review valve with releases >= MAX_RELEASES and gets STEP_CLOSE, cancelling the claim.

Worth a look

Nothing stood out.

---

_{Read: 2 files, +13 -5, the whole diff. Head 21642fe. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.}

[dmitrii-f-t27]

CI on 21642fe: all 15 Tests jobs and GitGuardian pass.

Code Quality is red. The findings are the same as on queen itself (run 38097722846 on 3c1b8c7, a recent push to queen):

• Biome: 20 errors and 122 warnings on both. The set of 126 file+rule findings is identical (diff is empty). The findings in queen-tick.ts and queen-dispatch.ts are existing noExcessiveCognitiveComplexity hits. Their line numbers moved by the added comment lines.
• Typecheck: the same 19 error TS* lines (file + code) on both, diff is empty.

This PR changes comments only, so it adds no findings. Merging into queen the same way as the recent merges into queen.

Dmitrii Fedorov (@dmitrii-f-t27)