First let me say that you should follow tilly's advice to
solve the problem. Having said that, you can
temporarily write a monitoring process launched via
cron to kill/restart the process if it reaches a threshold.
Check out Re: Re: HTTP Daemonology for details.
Notes:
The code (mine) is not optimized and can be improved.
Again, as tilly states, you should document the use
of the cron job.