the tack I would take there would be to disrupt the daemon w/ some kind of test harness e.g.

watch -n1 "pkill …"

and then harden the client connect to the point that it can reliably withstand active disruption like that