Much like the TraceEnable directive you added to the httpd.conf file I think having the RLimit directives added in could also be a good idea.

RlimitCPU
RLimitNPROC
RLimitMEM

CPU and MEM should be tailored to the users server based on the detected CPU and Memory (and arch).

Added to feature queue for feasibility research.

What sort of limits do you think are good?
RLimitCPU - say 70-80% ?

RLimitNPROC should not be less than MaxClients, and for things like SuExec and possibly the new php-cgi it may spawn more child processes, so something like (MaxClients + 20)?

RLimitMEM is per process limit so I can suggest to find maximum value of virtual memory size and then double it
( (ps h -U apache -ovsz | sort -rn | head -1) * 2 ) * 1024)
(you have to multiply by 1024 because it reports back in KB and RLimitMEM is in bytes)

Across all of my systems:
this was the highest number of an individual system 4464189440
This was the average of several dozen 2981020558

So a Good(?) proposed limit:
RLimitCPU 75
RLimitNPROC 535 (assuming max clients is 512)
RLimitMEM 3500000000


Thoughts?