1) When getting random computer errors, check the memory. Even if it passes a single pass through MemTest86, there may be timing errors between multiple chips.
2) When building a new system, always, always, always update the BIOS before installing the OS. The updates exist for a reason, and can save headaches later.
3) When in doubt, pull 1/2 of the memory in the system, run it for a bit that way, then pull the other 1/2. This eliminates sockets and controller paths from the problem.
4) A 24-hour heavy-duty burn-in is a MUST on any critical system. Forgetting to do this causes issues of "it ran fine during setup" and "what in the world is THAT?"
If I had just followed these simple 4 rules, I may not have burned a 1/2 tank of gas running to work after-hours for emergencies involving the new server since Tuesday...