This is a known issue on Debian/Ubuntu, at least. Sometimes the
pidfile is empty after startup, and this is why stop fails (it doesn't
know what to stop). This has been reported intermittently for a long
time now but no one has yet figured out why it happens. It's almost
always reported on ubuntu, but I think that might just be a measure of
its popularity than anything else.
Workarounds include using runit instead of the init.d script as well
as a manual kill -9 if the pidfile is empty.
It appears to only happen sometimes on first install or at system
startup and then seems to right itself for subsequent stop/start
calls.
B.