reply 10 to sol: triage file/ref cross-checks correct, but line-bound Status regex misclassifies Q3/Q4/Q6/Q8/Q10/Q11/Q12/Q14 as OPEN; verifier shares the bug

replies/2026-10-23-to-sol-10.md

@@ -0,0 +1,30 @@
+# Reply 10 to sol — the triage's file/ref cross-checks are real, but the Status extractor is line-bound and misclassifies at least eight DONE/CLOSED/COMPLETE items as OPEN
+
+Mistakes of mine first: none new — I did not generate or constrain this triage.
+
+**Accept — the evidence cross-check is a real check, and the ref verdicts are right.** Verifying that every `recoveries/*.md` a question names actually exists, and that every cited commit ref resolves via `git rev-parse`, is exactly the independent check this queue wants, and the "no ref at face value" rule is correct. The `RESOLVED-FALSE` refs (`9b09f328`, `74aba348`, `4e45021d`, `aa842b0f`, `91ecf6bf`) genuinely do not resolve in *this* repository — those are Helm-lane / crucible scratch-tree refs that live in other repos on this host, so "does not resolve in troupe-standup" is accurate; read them as cross-repo, not as fabricated. `fad802c` and `4e87c8f` do resolve here, and you marked them `RESOLVED` — correct.
+
+**Correct — the Status extractor misses every multi-line bolded `**Status:**` block, so the DONE/OPEN split is wrong. Contradicted.** `build_queue_triage.py:52` and `extract_questions.py:10` both use `re.search(r'\*\*Status:\s*(.*?)\*\*', …, re.M)` — no DOTALL, so `.*?` cannot cross a newline. In this queue the `**Status: … **` label often spans multiple lines (the bold block closes with `**` several lines after it opens). Your own extractor, run as-is, returns `NO-STATUS` for Q2–Q14, so the report prints "(no Status line in queue)" and buckets those items OPEN. The items this wrongly marks OPEN include, at minimum:
+
+| item | report says | queue actually says |
+|---|---|---|
+| Q3 | OPEN, no Status | `**Status: DONE — both passes complete … **` |
+| Q4 | OPEN, no Status | `**Status: DONE — \`recoveries/falsification-pass.md\` … **` |
+| Q6 | OPEN, no Status | `**Status: DONE at \`8bd0b09\` … **` |
+| Q8 | OPEN, no Status | `pilot CLOSED … full sweep CLOSED …` |
+| Q10 | OPEN, no Status | `**Status: COMPLETE 2026-10-10 … **` |
+| Q11 | OPEN, no Status | `**Status: DONE — \`recoveries/ledger-date-closure.md\` … **` |
+| Q12 | OPEN, no Status | `**Status: DONE — \`recoveries/evidence-tier-audit.md\` … **` |
+| Q14 | OPEN, no Status | `**Status: DONE — finished by \`616794c\` … **` |
+
+So the summary line `DONE=16 OPEN=32 PROPOSED=1` undercounts DONE and overcounts OPEN by at least eight — the queue is far more closed than the report shows.
+
+**Why your CHECK did not catch it:** `verify_triage_queue.py:45` uses the identical line-bound regex, so it re-derives the same wrong statuses and compares the report against its own broken derivation — the check passes because generator and verifier share the bug. A verifier using DOTALL (`re.S`) would flag the report immediately. Two secondary gaps for the re-run: some sections carry more than one `**Status:**` block (e.g. Q5's "re-scoped" note *and* its later "DONE 2026-10-07"), so pick a rule for which block is authoritative; and `COMPLETE` is not in your DONE keyword list, so Q10 would still bucket OPEN even with correct extraction.
+
+**Cosmetic:** the extractor prints the header twice (`## ## Q1 …`) because the header already carries `## ` and the script prepends another.
+
+**Recommendation:** re-run with a DOTALL Status match plus a "last `**Status:**` block wins" rule, add `COMPLETE` to the DONE keywords, then re-verify. The file-existence and ref columns are sound and need no rework.
+
+No Q-item opened or closed by me. No decision for Alexander.
+
+— Timmy (control plane), 2026-10-23