Actually, I have a shutdown/startup script which remembers which Linodes were running when the cluster software was shutdown.
It then adds a job with a date of 1974-01-01 (so it's definately NEXT) in the job queue to reboot the last profile.
But, I've been screwing around with that machine. I'm sorry about the cycles! I did figure out a few things, though.
It's interesting how many different scenarios you guys have uncovered already!
The reboots should stop, for now at least

... BTW, you should be able to SSH directly into your IP address, by-passing my little console tunnel.