tmux server dies silently causing orchestrator session crash loop

[zenprocess]

Summary

When running ao start, the tmux server dies silently after a few minutes, causing the orchestrator session to disappear. The dashboard (Next.js) stays alive on port 3000, but the orchestrator session transitions to "stuck" and eventually becomes unreachable.

Attempting ao start again then hits the stale port (related: #645, #300) and falls back to 3001, compounding the problem.

Steps to reproduce

1. ao start in a project directory (Linux, tmux 3.2a)
2. Wait ~5 minutes
3. tmux list-sessions → no server running on /tmp/tmux-501/default
4. ao status shows sessions as "exited" / orchestrator as "unknown"
5. ao start again → starts on port 3001 because 3000 is still held by the orphaned Next.js process

Expected behavior

• tmux sessions should persist for the lifetime of the orchestrator
• If tmux dies, the lifecycle worker should detect it and either restart tmux or cleanly shut down the dashboard

Environment

• OS: Linux (OrbStack, aarch64)
• tmux: 3.2a
• agent-orchestrator: main @ 4741ba2
• Node: v22.x

Related issues

• fix: auto-increment port when configured port is already in use #300 — port auto-increment
• ao stop fails to kill dashboard when port was auto-reassigned #645 — ao stop can't kill dashboard on reassigned port

[sreyom31]

Fix submitted in #696