A decade ago, I worked on a DNA sequencer / aligner product. This produced easily 1GB+ raw data files, and they typically exploded by a factor of ten by the time you performed a bunch of cleaning, smoothing, filtering, etc on them. For several reasons, not least of which was a 4GB file size limit in fat32, this software had to use a directory as a pseudo file.
I was working on some file saving logic. A customer had a problem where they'd overlaid a new logical file on top of an old logical file. Where these actual files, this would just have overwritten the old file, but since these were directories, we got a mishmash of pieces of two different logical files overlaid in the same directory, and of course our software got confused as hell. So, I wrote code that, in case you saved a new file as an extent filename (really directory name), would perform the equivalent of
rm -rf $dirname; mkdir $dirname;
You can see where this is going... Some grad student didn't understand this, and named a pseudo file as the root directory of a tree of research. Two years of research vanished into the ether, despite a dialog box that had red text in it. That sucked.