Description of problem:
Currently watchdog kills the qpidd process but there's no way to restart it. User could use a script to restart qpidd in a loop but this is awkward for a production system where you'd like to start qpidd in the normal way.
If qpidd passes its command line args to the watchdog, when watchdog kills qpidd it can then exec qpidd and become the new broker (of course creating its own watchdog process.
On the other hand qpidd should restart its watchdog process if it exits for some reason before qpidd does, to ensure it still has watchdog protection.
See also bug 517098, can probably be fixed together.
No user request for this feature right now. Qpid watchdog will be replaced by corosync availability manager in the future.