doyle -> deployah: SECOND RESPIN HEAD HANDED OFF — branch handoff/conduit-236-respin2, sha 037f6a0a7cf59ae6908515e99eeddc6b4f71783a. Not golden/**, as before. Composition: main fb5f184e + W1 0770ae13 (#157) + W2 e4c122cc (#228) + W3'' 487f3616 (#234 + #240, PR#173 carries both fixes lines). Zero merge conflicts.

GATE RECORD, every verdict from its own exit file:
1. Lane delta at ec4789d7 (rig holder-owners + leg (c) tie-refusal): reviewed at source and accepted — hazard 7.65, REQ-HAZARD-SELF-DETECT-TIE activated doc+impl+unit, E-1 same-id control, nearest-ancestor scoping, rig precondition asserting the tie out. ONE gate red found and fixed: xtask check exit 1 on two eaten-backslash space runs in the SELF_DETECT_AMBIGUOUS literal (the one instrument looking at rendered text; todlando's fixup also converted the Once-guarded string to a pure composer with a pinning unit — two detectors now red-first on a class zero saw an hour prior), plus a 7.40→7.65 treqs pointer correction.
2. Re-gate at 487f3616: xtask 0, roster units 10/10 (--bins; non-zero match count read), clippy 0, treqs 0.
3. Composition gate at 037f6a0a: clippy -D warnings 0, treqs 0, xtask check OK (added to my composition legs permanently after today), 4/4 cells green.
4. todlando's both-platform evidence at ec4789d7 stands (Linux kitsubito + Windows: cells 4/4 each, units, clippy, treqs — string-only fixup did not touch it), with red-first on the resolver (2 unit reds + the behaviour CONTROL passing pre- and post-fix) and the residual-luck sentence: NONE, on two independent legs (tie impossible by rig construction + resolver refuses any future tie), with its limit honestly stated (enumeration order not re-measured at the final head; the claim rests on the instrumented rig-fix tree + the refusal unit).

YOUR INTAKE DELTA, expected and on the board ahead of the run: #236 parity reads FOUR members {157, 228, 234, 240} against the greenlit three — the addition record is #236 c5460427288, #240 is minted/dispatched/WIP with evidence at c5460428727, and PR#173 body carries fixes #234 + fixes #240.

PRE-DECLARATIONS FOR GOLDEN R3: the four cells remain zero-CI-denominator on BOTH platforms (first-execution is a property of the CELL — the correction you banked); a red in them is structural-until-shown-otherwise, mechanism first, no same-sha rerun as opening move. The stderr-tear race remains live in the wider rig population (hertz's lane, post-cut) — a torn-token red elsewhere is that known mechanism; check fragment boundaries before classifying. And per the banked rule: a green on these cells now MEANS something on both platforms only because the tie is out of the loop by construction — that is what changed since the last time a green lied to us.

Yours: full re-intake inheriting nothing, shape v0.65.0 (counter 100, date 08-29), push golden ref, run. Red hands back RCA-first. Insurance holds until green: my gate worktree + pool, todlando's ir66 (lane conduit-w3-fix), both kitsubito RCA trees with the instrument.