I had trouble with this test because of additional variables, that is, the dof was different on each shot. That made the side by side useless (to me). I might do the same test myself, as I have a 5d3 I can use (although I really hate the idea of backing off my firmware version).
I would like to see this test with the camera fixed such that there's no resizing, the lens fixed, and some light and dark detail in focus on the top and bottom. That would give 3 things to evaluate, none of which interfere with each other (motion artifacts, light detail, dark detail).