rdmpw3275m-changelog-20260923-1011-jdmbair13m5-resume-and-fleet-close-out
rdmpw3275m changelog — jdmbair13m5 post-reboot resume, fleet Claude Code rollout closed out
[2026-09-23 10:07:52 → 10:11 EDT · rdmpw3275m · claude Opus 5.5 (1M) · session 5739b30e]
Rich: "resume".
Final state (verified 10:09:39 EDT)
- All 6 hosts on Claude Code 2.1.280.
- No top-level
model/effortLevelpin on any host. The resolved default isclaude-opus-5-5[1m]. permissions.defaultModeisbypassPermissionseverywhere.- All 36 live interactive sessions run 2.1.280: rdmpw3275m 6, rdmsm4x 7, rdmbair15m5 5, rdmbair13m5 5, jdmbair13m5 6, rdmpw3265m 7.
- The reconciler held the fix: configsync ran 24 times after the 2026-09-22 19:47 canonical fix without re-pinning anything.
Overnight / this morning
- rdmbair15m5
2a3bf406: the deferred daemon waited for the 22:52 check-in, then restarted it at 23:02. Result ok: v2.1.280, same session id, transcript grew, bridge restored. - jdmbair13m5:
- Rich logged in at 00:38 after the 19:56 reboot.
- The host's own updater then upgraded to 2.1.280 at 00:42, so the Homebrew hang did not recur.
- The new binary's first exec hung behind
syspolicyd, the 4th host with that wedge; it was detected and logged by claude_code_update.zsh v1.2. - The 09:05 reboot cleared it.
- 4 of 5 sessions were relaunched automatically at login.
3c2c5c6e, which was a bareclaudewith no id in argv, was relaunched by me at 10:08 in a background Terminal (open -g): v2.1.280, idle.
- The hub watcher/reporter
(
nohup … & disownfrom ssh) died silently overnight, and no reaper was found. Lesson recorded in memory (here and on the hub): use a double-fork daemon or a launchd one-shot.
Tickets
- Resolved: ISSUE-20260922-11, ISSUE-20260922-12, DEC-20260922-02.
- ISSUE-20260831-19: recorded the 4th recurrence and suggested the updater auto-restart syspolicyd on that signature.