You are right that gameports are very demanding. For some reason that little fact escaped me before. 😁 I believe the original gameport is the most wasteful. There was a brief "digital gameport" era just before USB came in, and that version of the gameport addressed the extreme CPU drain I believe. Most PCI sound cards have those gameports. Just getting it off of the ISA bus probably helped some.
If I'm using Win9x, I always try to go PS/2 for keyboard and mouse. Seems to be the fastest, least CPU intensive way to go I think. On modern PCs, I just stick with USB though because there's little to gain these days.
Back when USB was rather new, with say the 430/440 chipset series or Super 7 shit, I remember having a lot of weird issues with USB. Part of it came down to cards not liking to share the USB IRQ and the IRQ was shared with a PCI slot. Win9x was also at fault here because IRQ sharing is not so great with it. Also, occasionally you'd run into a mobo with one of those horrible early ACPI implementations and I'm sure weird stuff cropped up with USB cuz of that. So I'm a bit USB-jaded, especially if working with old hardware.