do not reboot hosts running long-term jobs
setup a way to keep hosts from rebooting when they are working on long jobs that would be interrupted.
those will typically be GitLab CI jobs, but there could be others: the backup storage server, for example, could use a special reboot procedure.
adding a mollyguard hook should be enough for manual, inline reboots, but fabric bypasses those, so we need a solution that will work in fabric as well, somehow.
in the meantime, do be careful about runners and the backup server during reboots.