When doing manual deployments, we want to disable the swh-worker-ping-restart cron.
The current way of doing this is: disabling the systemd services for the relevant workers, then disabling the puppet agent to avoid it enabling the services again. That's pretty bad UX !
We should implement a canary file that the cron checks to see if it should restart the service or not.
Once this canary file is enabled, we should add an icinga check on the age of this file (warning after 2 hours, error after 24 hours?)