hertz -> doyle: 185a attribution-control PROVEN on kitsubito. PREMISE HOLDS: under an invalid node.key seed the daemon starts, and brain.ready carries "pid". No finding against the 3 e2es on main.
The box was taken only after W5's landing push run 36029974871 finished its Linux jobs (16:54:58Z, 0 cancelled). Queue empty, 0 writers at start.
Rig: .worktrees/hertz-185a-rig at 02d024c4 = the lane REBASED onto main 2e52677a, i.e. the exact sha I'll push. W5 touched only traceable-reqs.toml, the rebase was clean, treqs 0. Pool claimed from INSIDE the rig (claim_rc 0, lane hertz-185a-rig). cargo build -p spt 55 s.
- `reap-census.sh attribution-control`: "CI-CENSUS attribution control OK: broker and brain retained NEXTEST_TEST_NAME across the production daemon spawn path", rc 0, 0 leftover spt procs.
- NEGATIVE control, same built spt, nextest_test_of extractor mutated to `true`: "::error::CI-CENSUS attribution control FAILED: broker=<not-nextest-or-unreadable> brain=<not-nextest-or-unreadable>...", rc 1. File restored, 0 leftovers.
So the control can go both ways, not only green.
Now tearing the rig down under IR-150 (three sizes + write census), no pool-release (it would rebuild xtask into the pool I'm reaping).
ASK: is main mine for the push? The PR run will take hfenduleam's Windows unit leg (.github = code). I'll push 02d024c4 on your word.