finding(plugin-grid): `ObjectGridDataTableSchemaHolds`' docblock forbids the exact declaration the 2026-08-30 ruling authorised — and the seam hold itself is now redundant

[os-sam]

Blocked-by: #6918

Filed by the domain:ui execution seat (PM session session_013hfmP9hoMd3dJwTh85J4yB) from the #6882 dev's report, at dispatch time rather than at merge time — see "why this is filed now" below. Unassigned and ungraded; grading and routing are triage's.

What was measured

packages/plugin-grid carries a seam hold, ObjectGridDataTableSchemaHolds, which exists because DataTableSchema did not declare renderCellEditor or schema-level cellClassName. Two things follow from PR #6918 (card #6882) declaring them:

1. The hold is redundant. It intersects keys the base type now declares itself.
2. ⚠️ Its docblock is worse than stale — it is actively misleading. It still states that the ruling is pending, and it carries an explicit prohibition against declaring these keys on DataTableSchema. The maintainer ruled the opposite on 2026-08-30 (batch Add default props to all components to prevent collapse in designer #4, verbatim 「同意」, option A).

⇒ The next agent to read that docblock will be told, in the repository's own voice, not to do the thing the maintainer has already decided should be done.

Severity: leaving it is not the neutral option

This is the defect class this repo keeps paying for, and three cards closed in the last day are instances of it:

• The record-drawer size: 'lg' bucket decision has no live tracker — three source files point at #6303 and #6259, both now closed #6584 — three renderer comments named closed cards as the open home of a decided question; a reader would have re-opened it as a cleanup.
• finding(components): the data-table's document-level pointerdown commit exists because "injected widgets have no blur handler" — field widgets now deliver a host onBlur #6859 — a comment justified a live listener with a claim that had stopped being true, and a later audit reasoned from it to a wrong conclusion about data loss.
• This card — a docblock that does not merely record a stale fact but issues a standing instruction contradicting a ruling.

The third shape is the worst of the three: a stale statement misleads a reader who checks it, while a stale prohibition stops them checking at all.

⛔ Why it was NOT folded into PR #6918

The #6882 dev declined to fold it in, and that was correct on two grounds this card records so nobody re-opens the question:

• Outside the ruling's landing surface. The ruling's stated surface is packages/types + docs + the one (schema as any) cast. Adding a published-plugin edit to a PR that reviewers were told is types + docs + cast changes what the clause-② review was scoped to.
• Not mechanically forced. The intersection DeclaredDataTableSchema & ObjectGridDataTableSchemaHolds still compiles; plugin-grid type-check exits 0 and its 183-file suite is green — verified, not assumed. So nothing is broken while this waits.

⭐ Why this is filed NOW rather than after #6918 merges

Because the alternative failure mode was measured on this exact repo, this week. #6584's own close-out states it:

A card that closes on Fixes # takes any deferred half with it. Where a dispatch deliberately narrows to the decided half of a question, the open half needs a card of its own at dispatch time, not at merge time.

#6584 lost its decision's home for four days that way. ⇒ This card exists before #6918 merges, so the deferred half has somewhere to live regardless of what happens to that PR.

Blocked-by: #6918 is in the body, not a comment — per #6653, 17 of 24 domain:ui blocked cards carry that line only in comments and are therefore invisible to the unlock scan's reverse index.

Scope, when it unblocks

1. Retire the ObjectGridDataTableSchemaHolds seam hold, or state in its docblock why it survives as a deliberate redundancy.
2. Correct the docblock either way — the prohibition must go regardless of what happens to the hold itself. ⚠️ If the hold is kept, this half is more urgent, not less.

⛔ Verify on the merged ref, not on this description: the hold's exact name, location and docblock wording were read on the #6882 branch before #6918 landed.