Day 5: crash. Day 6: crash. Each day, we were receiving notifications from Rackspace that the server had exceeded its memory allocation and was thrashing too hard; it rebooted our server for us.
My initial reaction: Increase the memories!
I looked deeper into the problem, and along with some colleagues, discovered that at midnight our server was kicking off 9 different processes at the same time. 9x Rails is just about 8x too many.