Daemon fails to start after being killed

[greschd]

When killing the daemon and removing the sockets directory given in circus-*.sockets, the damon can no longer start:

$ verdi daemon start
Starting the daemon... FAILED
Try to run 'verdi daemon start --foreground' to potentially see the exception

When called with --foreground it hangs at

$ verdi daemon start --foreground
Starting the daemon... 2018-08-10 02:13:53 circus[17615] [INFO] Starting master on pid 17615
2018-08-10 02:13:53 circus[17615] [INFO] Arbiter now waiting for commands
2018-08-10 02:13:53 circus[17615] [INFO] aiida-default started
2018-08-10 02:13:53 circus[17615] [INFO] circusd-stats started
2018-08-10 02:13:53 circus[17628] [INFO] Starting the stats streamer

I would guess the reason is that the daemon tries to connect on the inexisting sockets, which obviously doesn't work anymore.

It's not a big issue, since obviously it shouldn't happen too often that the daemon is manually SIGKILLed. However, maybe there's an easy solution to prevent this issue? When starting the daemon, can we make a sanity check whether the socket directory exists?