Ah, yes, the floating point bug that set the FP error handing to "software" instead of "hardware". Of course, no one thought to inform the application software side of the house, so you were one divide-by-zero away from an application crash because the FPU said "not my problem anymore". You could fix it with a liberal sprinkling of _fpreset(), assuming your language/environment allowed that option.

To be fair, HP wasn't the only one that had that bug. My SWAG is that there was some common library that a lot of drivers used, including HP, that fiddled with the FPU and never set it back.