Let a Copilot File Table Carry Forward on an Identical File Set

[ptr727]

The Merge Gate in pr-review-conduct carries an earlier round's coverage statement forward when the pull request changes the same set of files at both commits. It does not carry an earlier round's file table forward. A table counts only on the round covering the head. Copilot's lite rounds keep producing exactly that case, so a merge that every other gate item clears goes to the maintainer as a coverage question.

Measured

Three feature PRs into develop in one session, 2026-10-01. Each had the same shape:

PR	Round 1	Re-review on the next head	File set between the two heads	Outcome
#2265	lite, coverage=table, 12 files	lite, two rounds, no table, no marker	identical, 12 files	maintainer accepted round 1's table
#2269	lite, coverage=table, 1 file	lite, no table, no marker, "Approval recommended"	identical, 1 file	maintainer accepted round 1's table
#2271	lite, coverage=table, 4 files	lite, no table, no marker, no findings	identical, 4 files	maintainer accepted round 1's table

On #2265, one explicit requestReviews re-request produced another round with no table, the same result #1692 recorded for the marker. In each case pr_review.py status exited 45 with NO FILE TABLE STANDS IN, while the digest's own change-set comparison could have shown the file set unchanged.

The question

Should a round's file table carry forward under the same bound a statement does, meaning the pull request changes exactly the set of files the table names, at both commits? The alternative is to keep the current rule, where every such PR needs the maintainer's reading.

• Option 1: carry the table on an identical file set. It uses the bound the statement carry already uses. pr_review.py already computes the change set at both commits for the statement carry (carry_holds). It removes a maintainer prompt that has been answered the same way every time it was asked.
• Option 2: keep the rule. A table says which files a round read. It is weaker than a marker that states a count, and a re-review on new content may have read less than the earlier round did. The maintainer's reading stays the safeguard.

A partial table, or a round that states partial coverage, would still go to the maintainer either way.

🤖 Generated with Claude Code

[ptr727]

Answered by the maintainer on 2026-10-01: option 1, carry the table. A full-file Copilot table on an earlier round counts as coverage of a later head when the pull request changes exactly the same set of files at both commits. That is the bound carry_holds already applies to a carried statement. A partial table, or any round on the pull request that states or appears to state partial coverage, still goes to the maintainer. Implementing it means changing the Merge Gate text in pr-review-conduct (and the GOVERNANCE.md section it carries) together with scripts/pr_review.py's coverage reading and its tests.