fleet synchronously stops units during init

[bcwaldon]

1. start a fleet unit with a long-running ExecStop command (maybe /bin/sleep 60s)
2. kill fleet in such a way that it does not have a chance to gracefully shut down
3. start fleet
4. fleet will wait for the long-running ExecStop command to finish before moving on to normal operation

I expect fleet to trigger all of the systemd jobs to get the machine's state back to the intended state tracked in the registry. It does this, but it is doing so synchronously [0]. If your units take a long time to shut down, this will block fleet from joining the cluster.

[0]

fleet/agent/agent.go

Line 129 in e0506c9

a.um.Stop(name)