Scheduled supervisor update — not yet live at the October 5, 2026, 15:11 PDT review. The durable maintenance job was waiting for its compatible bridge to finish. Initial activation is authorized only after all four lobbies are confirmed empty and idle.
Five seconds is the eligibility wait, not a promise that the complete handoff finishes within five seconds. Relisting, match countdowns, pre-countdowns and empty-difficulty resets are separate timers and are unchanged.
File: /opt/poolrotator-runtime/instances/supervisor.json
{"deployment_empty_seconds": 5}
Accepts finite numbers from 5 through 3600 seconds. The existing supervisor tick hot-loads the file locally; no API polling or new monitoring thread is added. Invalid JSON, unknown keys, booleans, nonfinite/out-of-range values and symlinks retain the last known good setting. Removing the file or key restores five seconds. Changing the value restarts the continuous-empty interval. The effective value appears as deployment_empty_seconds in the operator rolling-status file.
The maintenance job waits for the existing rollout, installs the compatible bridge through empty-safe room updates, then waits for all four rooms simultaneously empty and idle. It confirms private gates and drains late events before stopping the old supervisor. A new supervisor starts only after closure and clean ownership are verified. Unknown closure or startup outcomes block for review instead of starting duplicate clients or retrying blindly.
Validation passed: 780 offline supervisor tests, 770 bridge tests and service-owner read-only checks of all four catalogs. These checks did not create live test rooms or use a second authenticated account client.
The owner separately hot-loaded the normal 90–135-second preference and 240-second hard cap in all four rooms on October 5. That configuration does not wait for v6.7.22. Jump retains its under-90-second preference; DT uses its existing shorter preferred minimum. Details and deployment protection.