When the USB MIDI driver was released, there were some reports of it failing, then working again, then failing. It was all quite curious.
I later found out that some other programs were causing the USB stack to stall, which obviously took the USB MIDI driver out too. As soon as the stall was corrected, the driver would automount the devices again.
I "solved" the problem by adding a requester whenever the USB driver would stall. This gave the end user a chance to understand what/why was causing the real problem. But only after experiencing it myself did I realize that I needed a timed requester, because if the USB stack has stalled, so has the keyboard and mouse...
Wow, that was a long story with a really boring ending. I'll have to work on that.
