Wow, this project looks amazing! I count myself among the long-time datahand users who have been getting increasingly worried over the past few years about what I'll do whenever my datahand gives out (may it last forever), so if you guys can figure out a way to build these things for a non-astronomical price, you can count me as interested in the possibility of a group-buy.
I don't in any way want to hijack or derail the thread here, but I figured I'd give my two cents as a datahand user about the changes to the original design I've always wanted to make (and might have attempted, if I'd had the money to buy a couple of spare units to mess around with). I've always thought the mappings were less than ideal, which is partly because I've got a Personal and not a Pro, but the tri-modal system they've got in place doesn't thrill me, and the fact that there are several keys that just send shifted versions of other keys makes software-remapping a nuisance (I learned to touch-type Dvorak, so I software-map my datahands to a Dvorak-like layout, which results in several keys being useless, like the one that gives a double-quote in the QWERTY mapping, and now produces a '{' in text mode). Having the ability to hack the firmware of a PIC to customize layouts would be great, especially given the ability to define our own modes and chorded keystrokes. I also always felt like the thumbs were under-used; the spacebar is maybe worthy of being the only key for the right thumb given how often it is used, but the return key doesn't deserve exclusive use of the left thumb's most ergonomic movement. Particularly given that the thumbs are the strongest fingers, adding a couple of smaller buttons under one or both thumbs might be a good idea (much like a number of the other ergonomic keyboard alternatives have done).
But the single biggest flaw has always been the mouse functionality; the built in system is simply useless. Not only is it totally inadequate in terms of speed and precision, but it also requires a mode shift, which makes it nearly impossible to do anything that requires a mouse and a keyboard at the same time. I use a trackball, but doing so means that either the right datahand unit (I'm right handed) or the trackball can be in a good ergonomic position; not both. I vary which one is in the good position depending on whether I'm doing primarily typing or primarily mousing, but it's hardly ideal. It also means that when I'm using the mouse, I only have access to half the keyboard. This is a problem for a lot of applications, including games, but also including graphical design applications; trying to learn to use Blender was nearly impossible. I actually had to get a flat QWERTY keyboard to lay on my desk and hen-peck with my left hand to use the basic interface; I gave up after about a week, which is too bad, as I'd love to learn modelling. I would imagine that other modelling and CAD programs would run into the same problem.
Everyone here seems to be working on the assumption that a trackpoint is the only option for improving the mouse functionality, but as a fan of trackballs, I've long thought that what I'd really like would be to simply have a small trackball (like the thumb-ball of the Logitech M570, for example) in the right thumb-well of the datahand unit. In the original datahand design, you would probably need to remove both the number-mode switch and the spacebar switch, but I think that given the ability to change the design a bit the spacebar should be salvageable (just by moving and/or reshaping it a bit). Combined with the ability to re-map things to our hearts content in firmware, I think this would result in better mouse-performance for the unit as a whole. Not only do trackballs have an advantage on speed and precision, but it should be pretty easy to have a key that, when pressed, caused the trackball to function as a scroll wheel (again, this is just some programming on the PIC to send the right scroll-wheel signals, since I'm pretty sure that scroll up/down are functionally just button 4 and 5 on most mice; it would just be a matter of deciding how many scroll-lines you map to a full rotation of the ball). If we re-jiger the keymaping a bit, we could probably even integrate three mouse buttons plus a "scroll" button into text-mode, so you would be able to actually use the mouse and be able to type all the basic letters at the same time.
Maybe other people really like trackpoints a lot, or aren't fans of trackballs, so I don't know how popular this idea is, but it's the one thing I've always, always longed to do to my own datahand (and probably would have attempted years ago if I could afford to buy a spare unit to experiment on). To me, the trackpoint would be essentially an improved version of the existing system; it would have the same basic flaws, just to a lesser extent. Introducing a trackball would require a larger re-design of the thumb unit, but would give a much better mouse-keyboard combo experience. There might still be applications where extreme precision is needed that a standalone finger trackball (e.g. L-track or Trackman Optical) would be better suited for, but I think an integrated thumb-trackball would perform much better in general use than a trackpoint for a number of usecases. For better speed, better precision, and scrolling, one thumb key would be a small price to pay. How hard this would be to do on the technical side, I don't really know; I'm really a software guy. I can do basic soldering, but beyond that I don't really have much hardware experience, and I've never done any CAD/CAM stuff. If you get to the programming phase and need help writing code for the PIC I'd be happy to offer my assistance, but I don't know if you'll actually need it.
Anyway, I will continue to watch your progress with quiet enthusiasm; keep up the good work, and keep us all posted on how it's going!