Hmm, seems like the biggest hurdle for end users will be the controller. Seems like a waste not to re-use the original since it would not require assembly other than reconnecting wires, and it does work with Soarer's Converter which is relatively inexpensive, easier to program, and I believe has more features. Has any testing been done looking at whether more scan codes can be generated using the a new matrix and the old controller? Doesn't really matter what the scan codes are as long as they are unique since they can be remapped with the Soarer's Converter.
Only other thing I would change is to make the navigation block area a regular 4x4 matrix with single 1x width keys with a capsense pad under each one. That way, The builder can install whatever they want there. Any unused barrel holes could be covered with a mask of plastic or other material. Alternatively to masking, a few different top plate configurations could be made available. Only difference would be how many holes are cut into it. The PCB would be the same.
If you look into making custom cases, the overall plate width could be reduced, and you could go back down to standard 1x width keys for the control block area also.
One last "wish" feature would be something like an erase-ease key, where the space bar would be split so that you could choose to re-assign either half to be a backspace key. You'd basically have to cut a Model M spacebar using a dremel or similar so that one barrel receptacle is on each side. Putting one more cap sense pad underneath that second barrel that is normally used to stabilize would make this attemptable at least. Such a mod hasn't been possible up until now because there is no sensor underneath that second barrel.
It would be a waste to not re-use the original, but unfortunately it is not possible.
Ok well it's possible, but it'd be really super difficult.
You see, the IBM controllers have a lookup table for the proper values of voltage (capacitance) change for each switch location. In order to re-use the original, we would have to make a PCB such that every single value is the same. This would involve a huge amount of time, money and prototyping, especially since the capacitive technology is from the 1960's, finding people to help with it is about impossible. We'd have to do it all on our own with no expertise. There are some people who, through lots of experimentation, have become somewhat accustomed to the ins and outs of IBM capacitave PCBs but it's still really tough.
Another problem is that the XT controller is not on a separate PCB from the rest. So you'd have to saw it off and then sand off the soldermask and solder to the traces. No fun. This might be an even bigger reason to not use the original PCB than the above.
The controller isn't really the hold up, in my opinion. There are currently at least two different replacement controller designs. A large GB would implement one of them on the PCB with a pick&place machine doing all the hard work. This would allow for a very easy time for kit builders. Even if the controller is on the PCB as it is for the original XT, it wouldn't be too bad, though it might be more expensive.
It's a great idea to have the nav cluster be a little more friendly, but it makes sense why it's not. This is because the PC/XT keyboard only has 83 keys. I believe this design requires 84, so a single "parts" keyboard can provide enough for 83 kits. If you fill up the area, then you will need to purchase two PC/XT keyboards as "parts", and you end up with a much lower ratio of "currently working" to "just another parts unit", which is a shame. I think having the option for more keys in that area is a good idea for people like me with multiple "parts" units already, but having it be the "standard" probably isn't great. Having different top cases will increase the price for everyone, but it will afford some more elegance in the design and more flexability. It's a good idea, and will have to be considered when the group-buy happens.
If you are keen on having a split spacebar, I would like to recommend getting a wheelwriter keyboard. The holes in the spacebar and "code" key line up perfectly with a standard IBM bottom row, replacing the "alt" key. You then have access to two thumb keys. The only issue is that you will have to move one of the stabilizer clips for the spacebar over a little for it to fully engage. THis is a very easy mod to do, only requiring a little glue and some time. I am a big fan of the split spacebar, and especially given IBM's excellent compatibility, think others shoudl look into it as well. Another option is to have a different bottom-row layout and option for multiple "code" keys, but it would require a separate PCB and plate for that layout, increasing cost.
Overall, you bring up a lot of good points. I hope you are satisfied with the explanation I've offered.