Changing macOS timezone invalidates the running daemon (supervisor_generation_stale)

[bierlingm]

This is a re-file of #879 under the discussion-first process. The original Issue was closed in the backlog sweep, not as a rejection of the bug. I hit the same failure today on 0.7.2 after a timezone change.

What happens

If the macOS timezone changes while the Prime Agent daemon is still running, the next prime-agent launch times out:

Error: Timed out after 30000ms waiting for the Prime Agent daemon response to "create".

The daemon log shows:

Supervisor command create failed: Error: Timed out connecting to daemon session worker: Error: supervisor_generation_stale

prime-agent status still reports the old daemon as current with 0 sessions, so prime-agent doctor --fix does not clean it up.

Environment

• prime-agent 0.7.2
• macOS, Node v26.5.0
• Daemon had been up since 2026-08-15
• System timezone changed on 2026-08-17 to Asia/Bishkek (/etc/localtime updated at 12:35 local)

Reproduction

1. Start prime-agent so the background daemon is running.
2. Leave that daemon up.
3. Change the macOS timezone (automatic travel update is enough).
4. Run prime-agent again.

Expected

The running daemon stays valid, or the client notices the identity mismatch and replaces it. doctor --fix should also treat this as a broken service.

Actual

New workers reject the supervisor as supervisor_generation_stale. create waits 30s and then the CLI crashes.

Root cause: getProcessStartId() uses ps -o lstart=, which is rendered in the current local timezone. After the change, the same pid gets a different start id. In this case:

• stored: ps:Sat Aug 15 00:10:51 2026
• observed: ps:Sat Aug 15 04:10:51 2026

Worker auth then fails the ownership check.

Workaround

prime-agent shutdown --force
prime-agent

A new supervisor records the current lstart and create succeeds immediately. A timezone-stable start id (or auto-recycle on mismatch) would avoid this.

[snimu]

Addressed by #1971