I see what you're getting at. With careful parsing of the ps output, you could know if you have
a working listener running or not. I would probably work that out, especially with command
line trial and error, as suggested earlier. That's in addition to my suggestion below.

More importantly, though, a dead listener waiting on a script to check it, is a problem with robustness.
A noticeable amount of time could pass before the problem is automatically corrected, and users could
see failures.

As the listener failure was said to happen in a week or two under current load (if I remember correctly),
I would schedule a cron job nightly to kill *all* fmxdbc_listener processes, zombies or otherwise, and then
start a fresh one. The brief lapse in FileMaker service would be extremely limited and therefore more
likely go completely unnoticed.