Why do you assume binning? That could be a difference between this new video-DSLR (capable of 4k but still Bayer interpolated) and the C 300 that only outputs 1080p but uses binning to increase image quality.
I was with you at the first sentence - the new 1D X shows that it's possible to sample some decently large sensors without resorting to binning. You lost me at the parentetical "(capable of 4k but still Bayer interpolated)" though: All these cameras will have to use Bayer demosaicing to get pixels, unless Canon is moving to non-Bayer technology.
What he means is, the C300 is an 8MP sensor, in the same way as my 7D is an 18MP sensor. It has 8 Million (near enough) photosites, 2mil red, 2mil blue, 4mil green. Same as my 7D has 18million photosites, 4.5mil red, 4.5milblue, 9mil green.
The C300 takes 1 red, 1 blue, and 2 green photosites and calls it a "video pixel", and there's 1920*1080 "video pixels on the sensor.
The 7D takes 1 red, 1 blue, and 2 green photosites, and calls it 4 pixels. The red pixel takes blue and green colour information from its neighbours, and so on. That's what he means by "bayer interpretation". The colour information of every pixel is only ever exact for one colour, and "near-enough" for the other two colours.
Technically, the C300 could (but doesn't) take 8MP stills photos exactly the same as my 7D takes 18MP stills photos. If they added some firmware (and probably processing power) to the same C300 sensor, they could take 3840*2160 video in the same way.