rdmsm4x-changelog-20260908-1816-notes-appleevent-intake-dead-restart
rdmsm4x-changelog-20260908-1816-notes-appleevent-intake-dead-restart
[2026-09-08 18:16:25 EDT · rdmsm4x] · session 158c9525 (notes-republish-pending-changelog) · work window 18:11–18:17 EDT
Summary: Apple Notes stopped accepting AppleEvents at 18:01:54 while its process stayed alive and idle, so every changelog publish on rdmsm4x hung to timeout regardless of sender (Terminal-hosted session, the launchd autopublish job, and the sshd route that had worked at 17:44). Restarted Notes with SIGTERM + relaunch at 18:16:01; it answered in 5 s and the autopublish job published the queued changelog at 18:16:13.
Scope
- Host: rdmsm4x. Process restarted: Notes.app (old pid 21542, launched 13:12:35; new pid 44766). No settings, files or other processes changed.
Evidence (measured)
- Peer dev-a9:
notes_changelog.zshfrom a Terminal-hosted session failed after 600 s; kickstartedcom.rdm.claude-notes-autopublishat 18:06:33, no run-complete by 18:11. - Job pid 81386 alive inside
notes_changelog.zsh(pid 3029), sending a 20 s probe every 30 s (osascript pids observed 18822, 26061, 28480, 33876). sampleof live probe pid 33876: main thread inAESendMessage→mach_msg(event sent, waiting for the reply).- tccd.system logged
TCCCreateDesignatedRequirementIdentityFromAuditTokenForServicefor pid 33876 at 18:14:22 (appleeventsd identity lookup happened). User tccd (pid 704) logged 17,465 lines in 10 min, none for kTCCServiceAppleEvents → Notes: the target-side check that every delivered event triggers never ran. Noauthorization_actionfor AppleEvents since 17:55: NOT the 2026-09-05 consent-prompt case. - Notes log: three
errAEEventNotPermitted … invalid right … port dead:1entries at 18:00:54, 18:01:24, 18:01:54 (the peer's probes, processed after each sender died), then nothing. Notes main thread sampled idle inmach_msgwith one reachability timer firing; 0.0% CPU. - Control: the same probe over
ssh richh@rdmsm4x(sshd holds the Automation grant, worked 17:44 per RTTy session) also -1712 after 30 s at 18:14. - Console idle 2,366 s at 18:11 (nobody editing in Notes). Boot 11:03:58 today.
Action and result
kill -TERM 21542→ exited in 2 s.open -ga Notes→ pid 44766. ssh probe answered80folders after 5 s.- The running job's last probe (deadline 18:16:33) succeeded: run.log
18:16:13 PUBLISHED rdmsm4x-changelog-20260908-1800-rtty-build502-fleet-deploy-appstore-readiness.md,published=1 … failed=0; ledger row present. - Independent verification over ssh at 18:18 EDT: in folder llmlog of
account [email protected]
(1,560 notes) the note whose name contains
rtty-build502count = 1, and it is the first note in the folder:rdmsm4x - changelog - rtty-build502-fleet-deploy-appstore-readiness…. Note the helper titles notes<host> - changelog - <summary> - <agent> - <task> - <project> - <when>; a filename-prefix query (rdmsm4x-changelog-20260908-18) returns 0 for every note the job ever published and is the wrong check.
How to tell the two Notes-publish failures apart
| Signal | TCC consent prompt (2026-09-05) | AppleEvent intake dead (this one) |
|---|---|---|
tccd authorization_action … Denied user-consent on
sender timeout |
yes | none |
| Notes log on each probe | errAEEventNotPermitted after sender death |
silence after the last dead-port error |
| sshd route / launchd job | works | fails too |
| Remedy | human grants Automation at the console | restart Notes; publish resumes |
Outstanding
- Ticket filed (see session) for the intake-death signature;
recurrence would justify a watchdog in
claude_notes_autopublish.zsh: after N failed probes with Notes idle and no tccd decision, relaunch Notes. - herdr is now 0.9.0 (path changed again): herdr panes have lost Automation again (HERDR-20260905-05). Not touched here.
Backups / undo
- Nothing to undo. Notes persists its store continuously; SIGTERM was accepted gracefully.