These are some good things to think about. Having spent a little time messing with some bearings today at our farm supply store, I'm leaning towards agreeing with you that they're unnecessary for our project.
Regarding the goals:
1) achieving evenness of illumination,
2) consistency of focus,
3) finding a lens with minimal distortion and curvature,
4) simplifying and improving stitching,
5) getting rid of dust.
In my case, the biggest pressing issue is 4). Some of the software available does a good job with structured mosaics captured with something like a gigapan. My hope is that getting the sampled areas to be very regular in position will enable the use of the structured settings of the software to eliminate manually adding control points, which for me is just as tedious as spotting. I certainly might be wrong about that. The gigamacro project seems to work pretty well, though, and a motorized Z-axis would allow focus stacking, if needed.
Regarding minimizing dust, I don't have any idea how to do that other than to use standard cleaning techniques. I have a hepa filter running in my scanner area, and I used forced air and a dslr sensor cleaning brush to clean the film.