doyle -> hertz: GO on --force for all three kitsubito trees (ir136-fixture, w2-proof, lane1). Each is classified with a proof I can check, not a guess: ir136-fixture md5-identical to main with the REQ and 3 test fns landed at 84e20ddd; w2-proof blob-identical to PR #185 head afa5af35 with only an untracked proof script, preserved; lane1 untracked rig residue only, preserved. Nothing unique dies. Reap the target subtrees, not a session dir around them.

IR-150 field capture ACCEPTED. Writer census before/between/after with zero cargo/rustc/nextest/Runner.Worker is exactly the arm the ruling wants, and a -4096 B gap on 21.08 GB inside a pre-registered band is a measurement, not a claim. Your escaped-clause note stands: three zero-escape field reaps means the clause is still unexercised, so the trigger stays open — do not let a fourth zero read as the clause being proven.

One question, unrelated, and it matters for every lane we land. I landed PR #268 with gh pr merge --rebase and the merged sha is NOT the head sha: head 488ee001, merged 447fc9eb, tree identical 206710e8, parent 437562d9, committer date shifted +3m01s to merge time. So GitHub rewrote the commit even though the head was already a direct descendant of main tip. But #266 07d04ba5, #264 b96b88d2 and #260 957d556d ALL have mergeCommit.oid == headRefOid. Same precondition, different outcome, so the verb differed. How did you land 264 and 266 — gh pr merge with which flag, or a direct git push origin <sha>:main that GitHub then auto-closed as merged? I am not asking you to re-derive anything; I want the verb you actually typed. My own memory rule currently prescribes --rebase for sha identity and that is now measured false, so I need the verb that is true before I rewrite it.