pr_review.py Passes a Clean-Verdict Copilot Round Whose Headline Sentence Names a Finding #2567

Problem

#2544's acceptance had two arms. #2562 implements the first, a verdict headline that is not the clean one while the open-findings count reads zero, which wait now reports as its own reading and exits 50 on. Decision #2564 shipped #2562 with that arm only and split the second arm out to this issue.

The second arm is a clean verdict whose headline sentence names outstanding findings, while the open-findings count reads zero and no thread or suppressed block is on the head. scripts/pr_review.py still reports such a round as a clean pass.

Constructed shape:

<!-- ccr-overview-v2 -->

### 🟢 Looks good

One minor wording nit remains in the retry helper.

**0 open findings**

Why it was split out

A clean verdict carries no structural marker that it names a finding, so this arm can only be a reading of the headline prose. #2564 records the measurement: across the second-format rounds measured on this repository, roughly a quarter of clean-verdict headlines named a remaining nit while stating no open finding and opening no thread, and nothing in the body says whether that nit is new or an earlier round's thread already answered. #2564 declined a prose heuristic on #2562 for that reason, since it would exit non-zero on most of those rounds and every Copilot wording change would need another reader change.

Done when

• A design is chosen that tells a clean-verdict headline naming a new, unthreaded finding apart from one naming an already answered thread, or this arm is closed as not worth a prose reader.
• Where a reader is added, a test feeds the constructed shape above, and a test feeds a clean headline that names only an answered thread and still reads clean.