I dutifully took a snapshot of the "oops" report and submitted a bug. The next thing suggested was to try to upstream kernel (2.6.x instead of 2.3.x?) which I followed the directions to install.
Now I get more crashes but less information from them. Occasionally the machine just freezes while I'm using X, so I can't swap to the text console and take a look. The recommended practice of using SSH and 'cat /proc/kmsg' didn't produce any additional output.
For a while it seemed predictable--- I would get a crash right after my RAID-1 rebuild finished. So I switched to the text console and I was able to see a crash message, while promptly scrolled off the bottom of the screen due to all the other alarming messages the kernel was sending out.
I finally got a setup that I believe would fix this: 'cat /proc/kmsg | head -60' in my text window. Unfortunately now the crashes don't occur as frequently. I gave up after a day or so and started using X again, only to get a freeze this afternoon. Maybe I should trigger a RAID rebuild deliberately... I think what's going on is that one of my hard drives is getting flaky, which triggers a RAID error, which triggers some bug in the kernel.