check-half-states: re-measure and re-pin `MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY` (188.3 of 2026-08-31 — the new alarm reads a factor of 2.11 on objectstack)

[claude]

Filed by the skills seat at the merge of PR #16406 (flight R), which gave H22's pinned closed-update divisor H8's staleness alarm. The alarm's first live run names this act.

The reading

scripts/pm/check-half-states.mjs pins MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY = 188.3, measured 2026-08-31. The alarm landed by PR #16406 reads, on objectstack-ai/objectstack at 19:5xZ today:

RATE PREMISE DRIFTED — this sweep observed ~397.9 closed-issue updates/day against a pinned MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY = 188.3/day, measured 2026-08-31 (7d ago): a factor of 2.11, outside the 2x band.

On objectstack-ai/objectui the same run reads OK (~175.5/day, factor 0.93). Flight N's earlier reading agreed (~312/day over 40 pages, ~414 over the leading 12).

What

The act the alarm names, by hand and not by a sweep (H8's rule — a pinned premise is checked against a sweep, never updated by one): re-measure the closed-issue update rate on objectstack-ai/objectstack over a stated window with the method the file's own docblock prescribes for the pin (rows read ÷ the updated_at span they cover, the horizon-driven closed pass), re-pin the constant to the measured value and its _AT to the measurement date, and state the window and the method in the constant's comment. The cost note in H22's census clause that reads the pin stays consistent by construction (it interpolates the constant). If the docblock prescribes a longer window than one sweep reaches, take the window it prescribes.

Acceptance

• node scripts/pm/check-half-states.mjs --self-test exit 0 with the case count unchanged or higher; the live sweep on objectstack prints the alarm's OK clause after the re-pin (factor inside the 2× band), objectui still OK.
• check-governed-merges --test exit 0 (not governed); whole-repo lint; the PR body quotes the before/after clauses.
• Pure code ⇒ the seat flips ready and arms SQUASH auto-merge at ACCEPT.

Refs #16393, PR #16406.

---

Generated by Claude Code

[claude]

Claim: PM loop round 8 (continued) — flight S, one pure-code card: the act the new H22 premise alarm names on its first live run — re-measure the closed-issue update rate on objectstack-ai/objectstack by hand (the method the ratchet file's own docblock prescribes for the pin, over the window it prescribes) and re-pin MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY (188.3 of 2026-08-31, read at a factor of 2.11 today) with its _AT, the window and the method stated in the constant's comment. Never updated by a sweep — H8's rule — so a hand measurement in a dev flight is the instrument. Unblocked by PR #16406 (flight R) merging at 21:07Z.
Session: session_019RfFHiRCSs3JXLK4cwcfox
Branch: claude/issue-16419-repin-closed-update-rate
Worktree: objectstack-issue-16419
Domain: domain:skills
File surface: scripts/pm/check-half-states.mjs (the pinned constant, its _AT, and their comment; the self-test only if a case pins the number). ⛔ No other file; no logic change.
Container & model: XS, mode:subagent, opus — dispatch-gates.mjs --tier: no path-derived mandate on scripts/pm/**.
Clause-②: no
Serial: no open PR touches check-half-states.mjs (PR #16406 merged 21:07Z). Pure code ⇒ in-seat review, then the seat flips ready and arms SQUASH auto-merge; skip-changeset. The PR carries Fixes #16419.

---

Generated by Claude Code

[claude]

os-dev-report

{
  "issue": 16419,
  "status": "done",
  "branch": "claude/issue-16419-repin-closed-update-rate",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/16423",
  "premise_still_valid": true,
  "summary": "Re-measured H22's closed-issue UPDATE rate BY HAND on objectstack-ai/objectstack over the endpoint the constant's own docblock prescribes (GET /repos/{repo}/issues?state=closed&sort=updated&direction=desc&per_page=100), one read of 12 pages started 2026-09-06T21:59Z: 400 rows / 0.964d = 415.1/day, re-read at depth 800 rows / 2.051d = 390.1/day and 1200 rows / 3.092d = 388.0/day. The three depths agree to 7%, so nothing had to be adjudicated between them; the pin takes the 400-row window, the docblock's own convention. MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY 188.3 -> 415.1, its _AT 2026-08-31 -> 2026-09-06, and the measurement block above them rewritten; the WHY-it-is-data paragraph and the updates-not-closures warning are byte-identical, as ruled. 14 self-test cases pinned 188.3 and were updated (named in the PR body and below); three prose sites that stated the old divisor in the PRESENT tense are corrected in place because this edit falsified them. The alarm is closed on this board and CI is fully green on the head.",
  "tests": "All at head 5d60e2aea5 (final commit; no commit after any run). HAND MEASUREMENT: 12 curl pages, HTTP 200 each, rows 1200 (570 PR rows), depths 400/800/1200 = 415.1 / 390.1 / 388.0 per day. CONTROL (quoted, never assigned): the read-only sweep observed ~388.0/day over the 12 pages its own closed pass read. BEFORE/AFTER, H22's clause on a read-only sweep of objectstack: before = '⚠️ RATE PREMISE DRIFTED — this sweep observed ~388.0 closed-issue updates/day against a pinned MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY = 188.3/day, measured 2026-08-31 (7d ago): a factor of 2.06, outside the 2x band.'; after = 'rate premise OK — observed ~388.0/day against pinned MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY = 415.1/day, measured 2026-09-06 (1d ago) (factor 0.93, band 2x).' objectui swept with THIS file after = DRIFTED, factor 0.42 (~174.9/day) — NOT the expected OK; reported as a finding below, not acted on. GATES (exit captured before any pipe): dispatch-gates --commands --repo objectstack-ai/objectstack derived 36 commands from the real change set, all 36 run, '✓ dispatch-gates --ran: 36 derived famil(ies) accounted for — 36 run, 0 NOT-MEASURED.'; check-half-states --self-test '✓ 2636 cases pass' (count unchanged); check-clause2-carriers --self-test '✓ 190 cases pass'; check-governed-merges --test scripts/pm/check-half-states.mjs exit 0 '✅ NOT governed — ordinary queue landing applies to a PR with exactly this file list.'; check-nul-bytes 'OK (scanned 8057 text file(s) ... no raw ASCII control bytes)'; whole-repo pnpm lint (eslint . --no-inline-config, NOT narrowed) through scripts/pm/os-verify-lock.sh slot issue-16419, acquired first try: 'VERDICT command-exit 0 · held the lock 70s · waited 0s'. Also ran the 5 roster gates whose list sits under a directory this path is in (published-list-mirrors + self-test, console-injection, engine-double-contract, i18n-stale-fill, pm-label-desc-cap) — all exit 0; check:published-readme-exports exits 3 = PREREQUISITE NOT MET (needs built dist for 37+ packages) = NOT MEASURED, unrelated to a diff touching no package. CI on 5d60e2aea5: all 35 check runs completed, 24 success / 11 skipped / 0 failure, including Lint & Repo Gates, TypeScript Type Check, Test Core (6 shards), Dogfood Regression Gate, Live half-state sweep, Check Changeset (skip-changeset applied), Governed Surface Queue Guard. No ablation was performed: this card changes a measured constant, not a guard, so the alarm's before/after pair IS the reverse verification and both legs are quoted above.",
  "mcp_calls": "0 — every GitHub read and write went through repo-scoped REST (probe 200 first); no MCP GitHub tool was called",
  "open_questions": [
    {
      "question": "The preserved warning paragraph says update events and closures 'differ by ~17x on this board' — true of 188.3 against the 10.7 closures/day measured 2026-08-31, and ~39x of the new pin against that same recorded closure rate. The self-test case computing exactly that ratio HAD to move (17.6 -> 38.8), so the file now carries two multipliers. The paragraph was ruled byte-identical, so I did not touch it.",
      "options": [
        "A. Leave it byte-identical (what I shipped) and let the seat decide — a new multiplier written there honestly needs a closure-rate measurement this card did not commission.",
        "B. Re-measure closures today and restate the paragraph's multiplier in the same PR — correct, but it edits a paragraph the ruling protected and widens the card.",
        "C. Drop the multiplier from the paragraph entirely ('differ by more than an order of magnitude') so it cannot go stale on any re-pin — durable, but still an edit to the protected bytes."
      ],
      "recommendation": "A for this PR, C as the follow-up the seat can rule in one line: the multiplier is a derived reading, and a paragraph that survives every re-pin is worth more than one restated every 30 days. B only if the closure rate is wanted as its own pinned datum."
    }
  ],
  "out_of_scope_findings": [
    "noted, not filed (per #16351: observation, not a reproducible defect / contract violation / metadata-authoring trap): one pinned divisor cannot describe two boards. Measured within the same hour with this file — objectstack ~388.0/day, objectui ~174.9/day, 2.2x apart — so after this re-pin the sweep reads objectui as DRIFTED (factor 0.42). Nothing breaks today: objectui's installed copy predates this machinery entirely (13,103 lines, no MEASURED_CLOSED_ISSUE_UPDATES_PER_DAY, no closedRateObserved), and when it is refreshed the alarm will say so on day one, which is the alarm working. What is stale is the adoption note (#11217) saying the only per-repo input of a verbatim copy is the anchor-issue variable; the divisor is measurably a second one.",
    "noted, not filed: the summary line's H22 clause carries a hardcoded '~2.1 days of update-recency' for the retired 4-page cap (scripts/pm/check-half-states.mjs, the census clause prose). It is a reading of the OLD cap at the OLD divisor; at the new pin the same 400 rows buy ~0.96d. Left alone deliberately — it is outside the ruled surface and reads as history in place; flagged so the next re-pin does not inherit it silently."
  ]
}

Generated by Claude Code

---

Generated by Claude Code