All four are landed in full. Their narratives live in COMPLETED.md; the plan now carries only Gamma and Lambda. Cross-references and VERIFICATION.md citations repointed.
Replace the self-comparing render-window loop with a genuinely discriminating
floor-vs-exact check; correct two stale claims; mark the Auto/Manual floor-parity
premise as unverified; drop the STARTPOS/ENDPOS comment's circular justification.
The millisecond floor lives in the custom-bounds field, not the engine, so
RENDER_BOUNDSFLAG=2 is now the only bounds mode: the two-position type, the
console verdict and the STARTPOS/ENDPOS drift probe all go. capture.cpp 697 -> 622.
Grid-ness of an edge was a proxy for "no floor could explain this count," not
the test itself — equal remainders on both edges cancel under a full floor. Now
checks all three floored models directly and corrects the SHORT/LONG
floor-signature docs.
Verdict can no longer print a false EXACT on an on-grid end, no longer names a
bounds channel a content-derived render never consulted, and the grid-align doc
premise is corrected without implementing it.
Batch captures now name the retained-render folder once instead of nothing; Auto/Manual tail modes refuse a 0-byte render like None does; VERIFICATION.md steps 1-3 no longer invite a false conclusion; docs/comments no longer overclaim.
The one-frame bound is not provable for a per-edge renderer; the test now shows where it breaks. Refused renders move out of the bank instead of being deleted, so the DAW experiment has something to read.
The tolerance is unchanged and now derived, not assumed: frameCountFor lands in {floor(L), ceil(L)}, so a non-frame-aligned window can never miss by more than a frame. Naming the source is what tells a self-bounding render from a short one.
daniel