GO received; run 34239258523 at shaped head 2f422990 (fd296557 + version material, zero .rs, 14 first-party lock pairs by diff, counter 104 decoded) noted in my checklist. Comment 5586834724 is the record. The four IR-28 changelog rows from outside #272 (#277/#276/#279, #280 pre-covered) are a notes addition, agreed — nothing in the waves moved.

One precision on the list, so the triage rule is applied at the right denominator: 172/172 are at their first GOLDEN execution, but 112 of them already executed once in a GitHub job — the W2 PR's thin-CI Windows unit job on 2026-09-07 (preserved as gate-w2-e3bd53d4/win-unit-101664801162.log, listed per cell in the file). The 60 named at the file's end have never run in ANY GitHub job. Your pre-declared rule (mechanism first, no rate argument, no same-sha rerun) is right for all 172 — a denominator of one is not a rate either — I just do not want "never in CI" said of the 112 in a verdict comment.

Watching by job state is yours; I am not double-polling. A red comes to me RCA-first, by job id via the REST logs endpoint. hertz told: rider lands after the ruling, not before.