Restart Recovery (normal restarts)

v2026.9.6: restart recovery helps unfinished conversations continue from saved progress. Use this page when you rebooted the host or restarted the Gateway on purpose. If an updater left the Gateway stopped, go to update recovery instead.

Official September notes still say: make and verify a backup before updating. Restart recovery does not replace that backup.

What you should see

  • After a clean Gateway restart, an in-flight task resumes from the last saved progress rather than starting from zero.
  • Long chats should reconnect faster on later 9.x builds (v2026.9.3 called out faster reconnects; 9.7 “smoother long conversations”).
  • Background sessions from v2026.8.2 can still be sitting beside the Control UI after the window closed.

If it did not resume

  1. Confirm the data directory was writable and not on a full disk. 10.9 will accept new writes after you free space, without restarting, once cleanup succeeds—it will not replay interrupted writes.
  2. Check openclaw logs for the session id you care about.
  3. Reopen the session from the sidebar; search and pin if the list is long (10.5 sidebar).
  4. If Doctor claims data is damaged after a Node/SQLite bump, that is the 10.9 stale-index path, not a normal restart. Update recovery.

Related