One thing that I found surprising was that I did not see the resolution to be any better than my 5D mark II on this video. It looks like the moire/aliasing is mostly gone, but in this specific scene where it was compared to a HDSLR (possibly to 5D mark II as the moire resembles more the 5D than that of 60D on that scene).
I would be fine with this quality of image, but at that price, I would have expected similar fine detail as RED. For comparison clip from Scarlet posted by somebody:
Looking the comments, it looks like it is not even 1080p but it still looks more detailed than 5D mk II or the C300, and actually in the executive scene the HDSLR looks more detailed than the C300 except for the horrible moire it has (and these details are not false details created by the aliasing because the aliasing false details show up mainly as horizontal lines (and combinations of them that cause rainbows because aliasing occurs before de-bauer) that should not be there.
Based on the examples from C300, I am leaning towards a conclusion that it might be possible to achieve similar image with 5D mark II by installing the Mosaic Engineering anti-moire filter on it. The anti-moire filter costs 300 which is not so big investment over the 5D mark II compared to the out-of-indie/prosumer-reach price of the C300.
I am waiting for 5D mark III, maybe it will offer some resolution improvements over the current HDSLRs *and* may fix the moire issue (if we are lucky). If that would happen, then the 5D3 will be a dream coming true and it will find its place on my "must-have" gadgets purchase backlog and will then receive higher priority than a new Mac. I hope Canon comes out with the 5D3 before Apple comes up with the next iMac refresh. Algorithmically reducing image size from 18Mpix sensor to 2K will result sharp images if no line skipping is used, but rather the pixels are calculated together. The image will be without moire and aliasing if it is done right and does not require any optical antialiasing filter. It requires lots of horse power for the ASIC that does the conversion, but I am hoping that this will be the case with DIGIC 5.