There are a number of ways forward with regards to what is going to live under the Raspberry Pi, but all of them have a current, active roadblock except for one.

I'm going to use both a Beaglebone Green and the Raspberry Pi for now. Comparing the eventual additional hardware costs given the low IO count, lack of an ADC, and only one (or two) hardware PWM pins available on the Pi, I think I might actually leave it that way.  There are two things which could still happen to change my mind, but I'll elaborate on those later.

Basically I want the smoother acceleration and ease-of-use that Machinekit gives me, but I also wanted to be able to run everything on the Pi. Software-based step generation does work, but not at the high data rate that I'm going to need to take advantage of how insanely low the mass of what my steppers need to throw around is.

If I can run the step generation at a higher step rate, I can then use the micro stepping of the drivers at a higher level.  Since the static load (the force trying to pull them one direction, in this case, gravity) on my steppers is so low, more microstepping may actually be somewhat accurate.  Accurate microstepping means I could run a slightly larger spool diameter and go faster. I'm not going to do that though as having the extra accuracy will allow me to push the edge of the build envelope out a little bit instead.  

With the tripod kinematics, achievable accuracy is directly related to your current position.  As you approach any cable, you start to trade accuracy in one dimension for accuracy in another.  The best balance of vertical/horizontal accuracy happens to be right were the boards for assembly are ending up, but where the feeders will be does trade some vertical accuracy for horizontal accuracy.  Having the extra steps, and if microstepping has a low enough load, the extra accuracy, will be useful and give me more breathing room there.

Moving the trajectory planning and step generation to the Beaglebone lets me use the PRU and therefore generate very accurate step trains at AM radio frequencies.. aka.. whatever rate I want.  That's, way, way faster than the driver chips could possibly go.

So the end result is I'm going to have two computers in there but that's fine, as the cost is about the same as adding all the little bits i would need to add without having it. The Beaglebone can deliver them all and then some.

Decision made, I wired up the Pi and Beaglebone, steppers,  basically everything I'm ready to mount yesterday.  I'll post some pictures later.