I was thinking about it some more, and I think it would be cheaper to just make a new PCB and controller than it would be to CNC a very complicated HHKB-like top-case&plate. Easier would be to laser a realforce like plate with nuts on the top for the screws to match up with (the screws that thread into the switches would stay, but the ones that screw into the plate standoffs would be replaced with a nut&bolt(&washer), slightly longer.
Then you could put it in an aluminium case, just like how the KMAC was done.
Yes, my idea is like harvesting parts from another keyboard, but instead of that, you are effectively "making" the parts you need. this means you wouldn't need a donor board, but you'd have to design and test more stuff. Much cheaper if many are made, but more work. That's why I think it's the worse solution, but it's different than what was mentioned already, so I thought I'd bring it up.
The reason I think it's worse is Moose's idea allows you to just use parts from one donor HHKB and requires less design work (and cost). Your idea of harvesting from a realforce / novatouch is more expensive, but is "known" to work well and also does not require time/design-cost/prototyping/etc.
But if we were to make this like the XTant project, then my idea becomes better I think. It only works for larger runs. Just a thought though.
3D printing might work. I think it'd be hard to get the proper tolerances required for low friction and so it wont bind on off-axis presses though.