The difference between 40 and 128 is 88. The is a 79% increase in resolution.
When we compare the 645D, with a sensor array size of 44 x 33mm , with the resulting stitch from the RhinoCam, with a virtual array size of 62.4 x 46.8 mm, there is an approx. sampling density
increase of +41.8% in each direction. However, sampling an MF lens with the RhinoCam closer to the limits of the image circle doesn't necessarily increase resolution at those edges as much as the sampling density suggests.
So a lot depends on the quality of the MF lens used, and especially its edge/corner performance. On the positive side, the dense sampling will extract the most resolution out of any given lens.
Also, the 645D doesn't use an AA-filter, and AFAIK the e.g. NEX does. I also do not know how well the NEX sensor responds to the more oblique rays at the more shifted positions. One may need to manually assemble a stitched LCC file (because Photoshop won't understand how to handle such light fall-off and Color cast) or perhaps it's easier to do it per tile before
stitching. The stitching should be able to handle the adjusted brightness differences between the tiles.
Naturally, you can stitch images taken with the Pentax 645D too.
Absolutely true and, when using rotational stitching, one always uses the center of the image circle of a given lens.