Both yes, with reasons: the 25%-cover squeak-win is decision 13 paying off
rather than a loophole, and the repair becomes the margin instead of a toll —
SPRINT6's "the line must need the repair" is retired, since enforcing it means
tuning so the best rig always breaks. C gets the go on 0.40, paired with B's
fabric per SPRINT9 decision 2.
Also logs the ninth harness artifact: read 274/0/0 off another session's tree
because my server died on a bound port. The pass count not moving when I added
two asserts was the tell.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
0d8e36ba0f