We are trying to figure out what's making the hardware so flaky. We need it to be able to stay up for at least an hour a few times so we can copy everything over to another box. The fsck takes forever all by itself.
Current theory is that it's something causing a short once in a while, as it's sometimes rebooting shortly after plugging keyboard, monitor or serial port in or out. In the past it'd reboot in the same way if we had a keyboard plugged in (but only then, we thought).
The perl.org MX is actually on another system, so it's (almost) only the mailing lists and the NNTP service that is affected. Mail to RT etc is generally still working.
I made a bunch of boot CDs. I booted it on memtest a couple of hours ago. Thursday morning (PST, thursday afternoon UTC) we'll go down[1] to the data center and see how the memtest went, if it stayed up through it and try to get it up and running one way or another.
When we've succeeded with copying the files I'll set the lists up on our new list server and the mail should start flowing again. Hopefully that'll be sometime late Thursday night or early Friday morning UTC.
[1] It's two floors underground, so literally down...
Steve Fink just reminded me that the tinderbox also used to run on onion. We'll try to get that moved over and setup again somewhere.
Thanks for your patience everyone. :-)
Good luck with everything. The effort is appreciated.