Why was it down, you ask? Well, it was out here in the lab, because of a shortage of network ports in the server room. From the acpid log:

[Sun Apr 15 18:53:07 2007] received event "button/power PWRF 00000080 00000001"

That is, at 18:53, someone simply pushed the power button. The server promptly closed all processes and properly shut itself down. I’ve moved it into the server room now…