hermes - 💡(How to fix) Fix Add command and auto-recovery for disabled dispatcher

The kanban dispatcher (running inside the gateway via dispatch_in_gateway) can silently become disabled when it encounters repeated failures (e.g., SQLite WAL corruption from concurrent access). When this happens, tasks in ready state sit indefinitely with no indication that dispatch has stopped.

There is currently no way to check dispatcher health without parsing journalctl logs. This is especially confusing because the board may have ready tasks that simply never get picked up.

Reproduction

1. Gateway running with dispatch_in_gateway: true
2. Kanban DB encounters corruption (e.g., WAL issue from concurrent access)
3. Dispatcher logs errors, then disables itself: "disabling dispatch for this board until the file changes or the gateway restarts"
4. Even after the DB recovers, the dispatcher stays disabled — requires a full gateway restart
5. No command exists to check this state

Requested features

A) Status command

hermes kanban dispatch-status

Returns:

{
  "board": "default",
  "enabled": true,
  "last_tick": "2026-05-26T06:05:00Z",
  "interval_seconds": 60,
  "ready_queue": 3,
  "last_error": null
}

When disabled: enabled: false and last_error contains the reason.

B) Auto-recovery

When the dispatcher disables itself due to DB errors, it should attempt re-enabling on subsequent ticks once the DB file is valid again, rather than requiring a manual gateway restart.

Hermes version: 0.14.0 (2026.5.16)