I recently had a problem with a rebooted PC throwing SCSI errors and the two attached hard drives resetting frequently until the thing just hung. I've used SCSI for a while and so my historically biased approach to diagnosing problems is "It's either cabling or termination". These days of idiot-proof automatic termination and blocks hard-wired to the cables reduces the chances of these kinds of events happening. The bottom line is, that after a great deal of jiggling and reseating, I found that the actual cause was that the SCSI ID jumper of one of the drives had worked loose and thus rendered it SCSI ID 0. Naturally the other drive already had that ID so the two of them were fighting. Maybe the smart move is to number them 1 and 2 so that unless they both shed their jumpers they'll remain unique. Anyway, I should have more quickly remembered the third historically biased diagnosis "...or bad SCSI IDs".
(Where will I be able to post my "I was an idiot" confessions if Advogato goes away?)