Motir

Filed by the motir run MOTIR-2942 of 2026-08-17. The correction is already applied — MOTIR-2939's description was amended on the record at 22:05Z. This is the telemetry, and the reason to file it separately from MOTIR-2942 is that it is a second instance of a shape whose first instance was recorded only as an aside.

What happened

MOTIR-2942's first acceptance criterion reads "the correction is already applied to MOTIR-2939's card and to the guard's comment block." The guard half was true — origin/main 45dd9a48 carries the per-commit table and UNCONVERTED_E2E_CEILING = 452. The card half was not.

The correction had been posted as an add_comment on MOTIR-2939 at 2026-08-17T20:50:49Z. get_work_item_activity MOTIR-2939 view: history returns ten changes and no descriptionMd edit since creation, so the description still carried:

• the two-row suspects table naming cloud-orb-clearance.spec.ts as a cause when it landed one minute before the baseline reading;
• the 4 / 3 per-spec counts, which count db.<model> statements the ceiling does not;
• and three plan clauses derived from them — "the 3 statements", "7 per-line swaps", "7 statements in 2 files" — against a fix that converted 5 and landed the ceiling at 452.

A reader arriving at MOTIR-2939 from the graph reads the description. Thirty-eight comments sat above the correction.

Why this is a SECOND instance and not the first

MOTIR-2907's close-out found the identical shape on MOTIR-2763: "⚠️ comment only. Its description still read '⚠️ Two surfaces' … A description is the spec; a comment is not. Written into the description during this close-out." That finding lives in a re-verification table row of a card about something else, and notes.html #292 — the entry it produced — is about a fix direction being unverified, not about where a correction lands. Nothing in either corpus has this as its subject, which is why the same run-shape reproduced it eight hours later.

notes.html #251 (MOTIR-2587) is the third member of the wider family from the other side — a re-measurement that amended the DOC and left the invalidated CARD untouched, naming it by key. Its lesson is about the card being skipped entirely; this one is about the card being touched in the half that is not the spec.

The generalisable half

An amendment is applied when the DESCRIPTION says so. A comment is a durable record of when and by whom, and it is the right place for the trace — but the description is what get_work_item returns first, what the item page renders as the spec, and what the next run reads to decide what to build. And the specific trap is who is standing there: both instances were made by the run that WROTE the correcting comment, for whom the comment and the amendment feel like the same act, and who is therefore the last reader who will notice the description is unchanged.

Corollary for this home: a telemetry card's "the correction is already applied to X" is a claim about X's DESCRIPTION, discharged by reading that description — not by having written the comment yourself. It is one get_work_item, and on both instances it was the check that was skipped.

Acceptance criteria

• No build work. MOTIR-2939's description is amended on the record (22:05Z, dated block, original table struck through and kept beside the measured one).
• notes.html carries the lesson as its own entry (#302 at authoring; locate it by CONTENT — parallel sessions append and the numbering shifts).
• Close when read. Promote to a plan-rules/ pack only on a THIRD instance — the corpus default is LESSON and run.md's calendar-check rule already says update_work_item ON THE RECORD, so the gap is a trigger, not a missing rule.

Context refs

• MOTIR-2942 — the card whose criterion asserted it; its close-out carries the verification table.
• MOTIR-2907 — instance one, recorded as a re-verification row.
• MOTIR-2587 — notes.html #251, the doc-instead-of-card sibling.