A manual script will definitely be better than running SIM if all you need is that done. SIM often breaks down the system (Atleast we've seen it happen). Then again for a full system work SIM (With well done configs) is handy. I think SIM overall makes sense because unlike the manual script, it will ensure httpd comes up with repeated tries (of course with the help of control panel and other system components).
But, would really restarting Apache due to a high server load, "FIX" things? Just seems like a temporary work around. Anyone have a good script for auto-suspending accounts that exceed a certain load?
This is correct - personally (and I've said this before in numerous threads about this same subject) I'd address why Apache causes runaway load instead of just putting a bullet in it's head every time it gets out of line.
As traffic starts to grow, here's what happens: sites start loading a bit slow; script kills apache; a bunch of users end up with half-loaded pages, so they hit refresh; apache gets hammered again, only this time it's funneled together more and during the process of restarting, which leads to more runaway load.
If Apache's configured correctly, you should not see high load averages. If you still do, you've outgrown your hardware - determine where the bottleneck is and upgrade accordingly. Magic cron-jobs to "fix" the problem will only make your service appear worse.