Oops, my bad. I restarted it Saturday morning and it appears to have been up since. I futzed with the ADB chain live a few days ago and that probably threw off the Rebound!
My memory leak is actually related to using the OT/SNMP extension which allows me to poll SNMP values (network, CPU, etc.) from the likes of MRTG. It is well known to have some nasty leaks. In fact, the box was very stable before I started polling it every 5 minutes with MRTG.
On all platforms, memory leaks are when you don't clean up old variables that are no longer needed when you're done with them. Eventually, you run out of free space in the memory space allocated to your application. More modern programming languages and runtime environments support stuff like automatic garbage collection to help developers with that task, but on the Mac OS it's all manual. Of course, other operating systems also have advanced memory protection so accessing memory outside your application's memory space (which, if you've accidentally filled yours up due to a memory leak, is more likely you'll accidentally do) won't crash anything but your app, but that can cause multiple apps or the whole system to crash under the Mac OS.
Mac OS X, being based on UNIX, has memory protection and if a developer is using Cocoa/Objective-C or Java they also have access to various types of garbage collection. Although, developers that use the oldschool Carbon APIs, they don't get garbage collection.
What do you attribute that to? The bigger L1 & L2 caches on the 604e & 604ev? The faster processor speed? I'd assume that you had the 50MHz bus speed in the 9500 if it was a stock 200MHz, so that'd be the same between the two models.
I only today was able to track down the AppleShare IP 6.2 updates (they're in "English-International", not "English-North_American" and not listed anywhere on Apple's support knowledge base that I can find), so I'll update to that since that's what's recomended under Mac OS 8.6. That may help increase stability, but I still attribute it to OT/SNMP.