I once fried an ipac hot-swapping from PS2.....
Are you sure it died from the swap and not a static discharge to one of the pins in the connector when you picked it up? That's a much more likely cause in that scenario. The lines going to the chip are the exact same ones used in USB mode, and it's only a different part of the firmware that gets executed.
You can do it a thousand times with no problem. It that 1001-th time that is the killer.
Yeah, kind of like driving to work, walking down a public street at night, or playing contact sports.

FWIW, I have popped a keyboard fuse in my past as well. But it was a P1 class Cyrix system, a big steel IBM clicky keyboard, it was a long time ago, and it pretty much did it on the second time I tried it. I also fixed it in about 15 minutes using the method described above.
This discussion is nearing being tantamount to people not going swimming in a pond, because their grandfather did when he was 4 and contracted polio from it. The early implementations were specced at providing up to 100ma through the port, so you can bet that the fuse they used was just a fraction over that (no hot-plug, so no significant overrating to account for current surges on the 5v line) The last PS/2 port fuse I looked at, which was on my "bottom of the barrel", 6+ year old emachines motherboard, was a self-resetting type rated at 3amps! That's 30 times more current than the port was originally specified to handle, and it doesn't fry. It acts like a thermal switch that opens when it gets hot and closes automatically when it cools off. The only real challenge is knowing just when they started building them better, so some caution may still be in order on old motherboards.
And just for fun, I have to share a moment of my own stupidity. I have a system in front of me that I develop on that has some cheap, side panel USB ports. I had the plug turned around the wrong way (I was working blind from the other side of the machine) and I attempted to put it in the socket. It didn't go, but it allowed enough of it to go in to short the connections. I now have two dead USB connections on that machine. Theoretically, that probably shouldn't have happened, but individual implementations will vary and therefore always allow for the occasional anomaly.
RandyT