At uni in the early 80s I managed somehow to end up specifying and then running the group PDP11/44, which was shared between a heap of postgrads doing signal and image processing (including me). 64kB of RAM, two 20MB hard disks and a tape drive, which was supposed to be used for regular backups but often wasn't.
IIRC the OS was RSX/11 which only had 2 categories of user, normal (restricted) and super (who could do anything) -- but to get access to the image acquisition hardware you needed to be a superuser, so most of the postgrads were. Late on day (Friday, obviously...) there was a major panic as everyone's VT100 stopped responding to commands, including directory listings. It turned out one idiot had decided to clean out his (multiple) directories, and entered RM [*,*]*.*;* (or the equivalent) and deleted not only all his own files but also everyone else's, and also the entire OS which is why no commands worked any more... :-(
(and of course there wasn't a backup HDD since they'd both been used, and it turned out that the last tape backup was several months old...)
In a stroke of luck the tiny little system debugger was sitting in memory because I'd been using it to try and track down a problem, and this quickly showed that all the HDD files had been deleted -- but the FAT was still there undisturbed since the disaster stopped anybody doing anything, just every block on the disk had been marked as unallocated. The debugger allowed the FAT (which included block number, filenames and used ID) to be inspected, and fortunately it could also drive the line printer directly since even the print command had gone.
Which meant I could print out the contents of the FAT, hand it to the miscreant, and tell him to manually reallocate every block on the HDD, which included the OS. IIRC there were 20480 1k blocks -- almost all used, which was why the attempted file clearance had happened -- and this took him all weekend, since the debugger could only work on on block at a time. Amazingly pretty much everything came back undamaged...
A close escape, all the PhD students would have lost months of work if not for the lucky fluke of the debugger still running.
There still wasn't any way to stop this happening again, except giving everyone an absolute b*llocking -- oh yes, and backing up to tape regularly...