Increases the delay before Swarm restarts a failed uptime-kuma task
from the default 5s to 30s. Applied live via `docker service update
--restart-delay 30s` on 2026-07-26; this persists it so it survives
the next stack deploy.
Note: this does not pin the task to a specific node or prevent
cross-node reschedule races against the CephFS-backed SQLite data
dir - it only slows the restart-after-failure loop. See prior
incident notes for the DB corruption root cause discussion.
admin
61b1a645bb