Some nice stuff, but I have a couple of disagreements.
1) He judges the "same FOV wide angle performance" by hobbling the smaller format with a lens designed for wider FOV on 35mm format and then heavily cropping that image on the sensor (he uses the Canon 24-70/2.8 at 24mm, whereas the other samples use that lens at longer lengths, where it can be expected to perform better). It should be well known that correcting for such wide angle coverage as 24mm in 35mm format hampers lens performance, and that is the specifically wide angular coverage needed that is the source of the problem.
A fair comparison would use a high quality zoom lens designed specifically for the smaller frame size, like the Nikon 17-55/2.8 used at 24mm; that lens only needs to be corrected for the same angular FOV as in his larger format examples.
At most, his comparison is showing that Canon's smaller format cameras are at a disadvantage at wide angle FOV because Canon is not providing top quality wide angle lenses specifically for them. (On the other hand, if the 17-85 EF-S is of suitably high quality, it should be used on the 20D for such a comparison.)
II. He is wrong about hand-holding shutter speed becaue he forgets the fourth factor effecting the degree of motion blur seen on prints: the degree of enlargement from the image on the sensor to the image on a print. With the same shutter speed and focal length in a smaller format, prints of the same size require more enlargement, making motion blur more visible. It indeed does make some sense to increase needed shutter speed in proportion to this extra degree of enlargment, which is the "format factor, or 1.6x in his example. This is the same as using the "35mm FOV equivalent focal length" in guidelines about shutter speed.