This is related to bug 185923. The problem is that when the fail2ban dies and leaves the socket file, it can't be started cleanly in the default configuration. Suggested solution is to alter the /etc/conf.d/fail2ban file, and OK, I won't argue about that. But please make some message either in the init script when start fails and the socket file is detected, or in the ebuild. Why it's so important? I had a few situations when fail2ban was not running. That's not nice, and as an admin I would like to prevent situations like that and know in advance about the -x option, not when something goes wrong.
(In reply to comment #0) > Why it's so important? I had a few situations when fail2ban was not running. you're telling that a longer error message would have avoided that situation? My opinion is it would have not changed anything. For whatever reason, the developers has chosen to not start fail2ban if the socket still exists, in the default configuration. i would rather suggest making "-x" the default configuration as from 0.8. This should not do anything harmful since, from the 0.8 version, fail2ban-client pings the server to guess if there is already one running. The /etc/conf.d/fail2ban config file is provided by upstream. I suggest you to contact cyril.jaquier@fail2ban.org