This reminds of me of the failure-mode-within-a-failure-mode of 10b2 with vaxstation2000's using vms's vaxcluster software. Unplugging the 10b2 gave you a window of about 10 seconds before one by one every vaxstation2000 would bugcheck. I was always rather astonished that nobody at DEC either noticed it, or thought it was a very big deal because the bug survived a long time.
I thought that was just me. My first IT job was developing credit-card systems on VAXen. We had the office flood-wired with 10base2 in one long bus - at locations where there wasn't a PC yet, there was just a faceplace with two BNC connectors, and a tiny patch lead between them. To install a new PC, you had to have a length of co-ax long enough to go from the faceplate to the desk and back, with a T-piece in the middle. Installation involved whipping out the short patch lead and re-connecting both ends of the longer one before things elsewhere declared the network as broken and started shutting down somewhat ungracefully. This was best done as a two-man job, but we did get it down to quite an art. Nice to know after all this time that someone else was playing the same silly game... Regards, Tim.