Restarting couchbase service when existing service is "stuck" causes problems and beam crash without user knowin

To reproduce:
Start Couchbase node normally
Use kill -STOP <pid> on both beam.smp processes to simulate stuck process
Run: /etc/init.d/couchbase-server restart
Note that the "stop" says failed, but the "start" says okay:
[root@cb1 couchbase]# /etc/init.d/couchbase-server restart
Stopping couchbase-serverNOTE: shutdown failed

Starting couchbase-server [ OK ]

There is now an erl_crash log present in /opt/couchbase/var/lib/couchbase.

To the user, Couchbase should not have been able to start if it was not able to stop properly. I've seen the same behavior except where the Stop doesn't not show an error, leading the user to even more confusion.