In fact the resolution of the F6.3 shot, when opened on my computer, is only 17MB, less than the file size of a 6mp camera, never mind 24mp.
The pictures with shallow DOF were just pictures taken with the camera - not intended to show corner-to-corner performance of the 24-70CZ on FF.
But the f/6.3 building shot was SPECIFICALLY shot to show corner to corner performance on Full-frame, with objects in the very corners (of 35mm FF) that can demonstrate the sharpness of the image at the borders. Actually the full-size JPEG was around 30+MB in size and had to be downsized but I would be very surprised if anybody were to make statements that they saw any kind of softness in it (there was NO softness, either at the borders or in the center, as an FYI )
This f/6.3 picture was specifically picked (among a bunch of others with less DOF or shots taken at an angle, which could be misleading) to counter the "internet rumor mill" which will tend to propagate speculations about performance, based on some shot somebody saw in some website. I think dpreview should have shot some brick-walls to demonstrate border resolution of the lens, than shooting images with a lot of depth or with variable depths once one gets to the borders, which will in turn be seized by people who are out to prove their own pre-set decisions.
Bottomline, the lens is SHARP, corner-to-corner on FF according to the shooter/beta-tester this person (who is also a Nikon D700/D3 shooter) was more impressed with the corner-to-corner performance of the CZ on A900, when compared to the new 24-70 f/2.8 Nikon version, which in turn is rated better than the Canon 24-70 f/2.8L (and this corner-to-corner rating has nothing to do with the Alpha/CZ version being Stabilized, while the others are not) . Does this provide any added talking points ?