todlando -> doyle. FRESH READINESS RECEIPT, attempt 2. Measured 2026-09-12T20:21:48Z-20:23Z, all axes re-taken now. Nothing launched, no harness edit, no elevated action, no rebuild.

1. PINNED BUNDLE -- INTACT. bundle-r4, all 9 files verify OK against BUNDLE-R4-FROZEN.sha256, byte-unchanged since the freeze.
   INSTRUMENT NOTE, because my first reading was a false red: plain sha256sum -c reported all 9 FAILED open or read. The manifest is CRLF and msys sha256sum takes the trailing \r as part of each filename -- 29 CR bytes measured with od. Re-verified against a CR-stripped copy: 9/9 OK. The bundle was never in question; my reader was.

2. PINNED EXECUTABLE -- UNCHANGED. sha256 edd3d8e0566690493182a0e1807b1a6b43c52cff45511b9e81490e256dd10e21, 38411776 bytes, mtime 2026-09-12T10:36:11Z. Identical to the 11:10Z census on all three.

3. SUBJECT SHA -- UNCHANGED. Worktree 304-w2-repr HEAD 53d625cd0bd88a04815efdf6c8209a3096bf53e8 == SUBJECT_SHA, porcelain 0 lines.

4. CAPACITY -- OK. 121 GiB free on C: against the 96 GiB floor.

5. FRESH HOME / EVIDENCE PATHS -- FRESH. fp-run and fp-home both ABSENT. Positive control run alongside (a path that must exist read as present), so the absence is measured, not a dead check.

6. LOCAL CENSUS -- NO BUILD CONTENTION. Zero cargo, zero rustc, zero Runner.Worker.exe. Runner.Listener.exe 22712 and RunnerService.exe 19588 are up, both since 2026-09-10T19:25 local -- so the runner can accept a job and none is running. 20 spt.exe processes are live infra; every one with a readable path runs from AppData\Local\spt-core\bin, NOT from the worktree target, so nothing holds the pinned executable. No pool claim is needed: S3 consumes --artifact and does not build.

7. GITHUB AVAILABILITY -- UNSTABLE, AND I AM NOT REPORTING THE CONVENIENT READING. Three probes, same predicate, inside about 30 seconds:
   probe 1: hfenduleam OFFLINE busy=false, kitsubito OFFLINE busy=false
   probe 2: both ONLINE busy=false
   probe 3: both ONLINE busy=false
   The local runner processes did not restart between probes (creation 2026-09-10, unchanged across all three), so a local bounce does not explain it. I did not determine which reading is authoritative or why they disagreed -- that would be investigation you have not asked for. Current state reads available on 2 consecutive probes; one sample 30s earlier said otherwise, and I am not treating 2 of 3 as a stable green. If attempt 2 depends on runner availability at a specific moment, re-probe at that moment rather than trusting this line. CI itself is landing: the last three runs on spt-bs-core all concluded success, most recent 2026-09-12T05:46Z.

RISK I AM FLAGGING RATHER THAN ACTING ON: bundle-r4 lives in the scratchpad of session 59484d41, which is dead. Scratchpads are what a capacity sweep reaps, and the freeze record names that path. It verifies clean right now. Copying it somewhere durable is not mine to do unasked -- say the word and I will, without touching the original.

Everything in this receipt is measurement. Attempt 2 remains ungranted and unlaunched; procedure v11 frozen at 7caa6b12; executor is liam per Reavo's restoration, session 1212783f-c73e-4b6f-a6b2-fa9a428720d5.