So, I recently decided to give Win2k another go on my NEC PC-9821 V166 after I gave it some aftermarket upgrades, like a K6-2 533, GA-VDB16/PCI (3dfx Banshee), and a 32GB SD card. I've already used it before in 2019 with the stock hardware configuration (P-MMX 166, 64MB PC66, Mystique MGA-1064SG), and while it did work it was unbearably slow. A known issue on PC-98 is that Win2k doesn't support the older generation of keyboards that had mechanical locks for the Caps and Kana keys instead of the newer software locks indicated by LEDs, just like modern keyboards, and as such doesn't even recognize it's presence. On the aforementioned first attempt I simply plugged in a USB keyboard and it did the trick.
This time however I'm not able to use any type of keyboard. For whatever reason, with the stock drivers and INFs, HID keyboards are recognized as HID mice. On the Japanese message boards I browsed USB keyboard problems weren't out of the ordinary. A common fix was to replace keyboard.inf and input.inf with their PC equivalents and then reinstall the keyboard driver, making sure to point to the I386 directory instead of NEC98, so that the necessary .sys files get replaced with the PC versions as well. This fixed the incorrect identification and the keyboard is now seen as an HID keyboard device... but it has a vague code 31 error. Do keep in mind that USB keyboards work perfectly fine during text mode setup, but in GUI setup and a fully installed OS they stop working. I even went as far as to plug it into a regular PC running Win2k, extract all the Enum registry keys, and transplant them to the PC-98, but that didn't work either.
Attached below is a screenshot I did through a TightVNC session (without which I wouldn't be able to type) showing the error:
I'm wondering if there's anything I can do by enabling debug mode and attaching WinDbg over serial. I'm not too well versed with it, but I'd like to at least do some deeper analysis in kbdclass and kbdhid.