"I'm not an expert on resolution, but 1280x1024 is in between 1080 and 720." It would be if more than 720 pixels of vertical resolution were used (e.g., if a 2:1 anamorphic lens and a 2.40:1 aspect ratio were used). Otherwise, at a 16:9 aspect ratio, the image would be cropped to 1,280 x 720, the same as 720p.

But would it? Is it possible that the principles of lens optics still apply, but at a much smaller size? I can see things like depth of field becoming less shallow, but possibly even not then. DoF is in essence a ratio. If all dimensions will be smaller, might they cancel out and still yield comparable DoFs at comparable effective focal lengths?

The DXL isn't using Red's color science, which is interesting.

One billion dollars.

Of course it's the exact wishlist (well, not exact -- should be global shutter with at least 15 stops of DR). Why spend millions to develop yet another camera that doesn't do everything on the wishlist? Even if they hit the specs they've listed, the wishlist is a moving target. Unless they're done most of the development, they probably won't release in December. By the time they do, and fix all the boo-boos, it may be too little too late. Nonetheless, I wish them well.

The sensor in the Ursa Mini 4.6K appears to be a custom-tweaked version of the Fairchild Imaging LTN4625A. The LTN4625A has a readout noise of <2 e- (root mean square [RMS]) in rolling-shutter (RS) mode and <5 e- (RMS) in global-shutter (GS) mode. It's RS dynamic range (DR) is >88 dB, or just shy of the 4.6K's 15 stops (90 dB). The equation to calculate DR is DR, dB = 20 x log(full-well capacity ÷ noise). If the RS and GS noise is the same for the 4.6K's sensor, then:
20 x log(full-well capacity ÷ 2) = 90
log(full-well capacity ÷ 2) = 4.5
full-well capacity = 2 x 10e4.5 = 63,245
The GS DR would be:
20 x log(63,245 ÷ 5) = 82, or 13⅔ stops

Of course, this is a theoretical figure, but it indicates that the loss in DR going from RS to GS might be 2 stops or less.