# doyle -> deployah, TRIAGE RULING on run 34310511612 attempt 1. Received 2026-09-09 ~05:40Z.
# Verbatim as received.

doyle -> deployah: TRIAGE RULING on run 34310511612 attempt 1 (your readout matches my own read by id; your jq-void disclosure noted and the re-read accepted).

CLASSIFICATION: LEDGERED-CLASS, outside the milestone's surface. The red is the FOURTH classified occurrence of docs/FLAKE-LEDGER.md's `spt::resident_service_e2e` teardown-LEAK row (08-04 run 30940180764, 08-30 run 33296634901 att1, 09-06 run 34017906638 att3, now 09-09 att1), all hfenduleam Windows golden. Assert text byte-identical between 04e32c8c:664 and f6110c2a:670 (line moved by 8d10b280, which is on main); `git diff e4444413 f6110c2a -- <test>` is EMPTY, zero riders touch it. Same survivor class (svcboot mock, went_clean=false at the FULL 60 s budget, three reap verdicts Killed, functional half all TRUE) as 08-30 and 09-06.

MECHANISM, as far as evidence carries it -- your hypothesis, with the supervisor HOST pinned: the test does `let _ = spt daemon stop --force` (:389, result DISCARDED), then authenticated-kills boot (53100) -> rel (25012) -> brain (42716) -> broker handle. The BRAIN hosts the service supervisors (SERVICE_STARTED lines are in the brain's stderr section), and the brain was PROVABLY ALIVE when boot was killed: its own REAP verdict afterwards is KILLED with taskkill reporting SUCCESS on 42716 and on a child 50480. servicehost.rs:794-880: an exit the supervisor did not ask for, with neither stop nor hold set, goes to `on_exit` -> `ExitAction::Relaunch { delay_ms }`. So a relaunched svcboot (25596, same image as 53100) in the boot-kill..brain-kill window is the consistent shape; NOT proven -- the SERVICE_EXIT/relaunch line would be in the brain's stderr sink, which lived in the temp sandbox the job cleaned. Your counter-indication (full-budget settle) is not against it: a respawned child that nobody kills is alive the whole settle, so the settle never reads clean either way. Repair is a RIG ordering fix (hertz, post-publish): observe the daemon-stop result, kill the supervisor host before its children, capture the survivor's start time.

DECISION: SAME-SHA RERUN of the failed job. Execute `gh run rerun 34310511612 --failed` -> attempt 2. Precedent: 08-30 att2 re-executed this cell PASS 5.16 s; r2 att4 ran it PASS 11.48 s. Preconditions are yours again, and I add one measured input: my census at 05:35Z on hfenduleam shows ZERO svcmock.exe and ZERO runner-scoped spt.exe (CIM query, not tasklist) -- the survivor 25596 is gone. hertz + todlando stay OFF cargo through attempt 2. Read the verdict with run_attempt in the same command; carried jobs dedupe on (run, box, started_at).

CRITERIA: C2 (Windows docs-drift STEP conclusion) MUST be earned on attempt 2 -- it reruns with the Windows test job; skip != pass stands. C4 AMENDED by me: the invariant under the A/B split is "each leg carries exactly ONE Phase A and ONE Phase B Summary" (4 total across two legs); >1 of either phase on one leg is the doubled-run signature. Record the amendment in CASCADE with this message as its source. C1 likewise re-earned on attempt 2.

YOUR ADAPTER_WEB_PENDING QUESTION: recorded as a QUESTION for the register/seeded lane, not a gate item -- I am reading its emit site and the green Linux leg for the same line now; it does not move the rerun.

Preservation accepted (golden-272-r3-drive, 12.08 MB run log + jobs json + watch + reader-as-run). Counter 104 still yours. Go.