I think it will be high frame rate because high frame rate is what Canon does better than anyone else. Not disputing that Canon does well with high framerate, but there are plenty of examples of Canon successes that have not been focused on speed. 5Dmk2 being one example. If anything, recent history seems to suggest that Canon should prioritize video functionality instead of still images, and one could argue that is exactly what they are doing :-(
I see no reason why that technology could not be applied to a "measly" 30-40mp FF sensor to achieve at least 6-8fps. I also see no reason why ISO range would have to suffer. High ISO capabilities are not mutually exclusive with low ISO capabilities. On the contrary, high ISO is limited by physics, while low ISO is limited by electronic noise sources. Canons maximum well capacity is already more than high enough to fully exploit 14 bit data, as well as fully exploit 16 bit data...the only thing in the way is their high read noise. That could be solved with a parallel digital readout approach that applies digital noise reduction similar to Sony. If Canon solves the noise problem, they could easily have both quality high and quality low ISO performance.I think that the pixelrate (sensor resolution times framerate) has some relevance to cameras. Internal bandwidth scales with pixelrate. Image processing dsp scales with pixelrate. Total camera heat output / battery drain probably scales pretty well with pixel rate.
It has been said that the differences in quality between my 7D and the 60D could be connected to the 7Ds higher framerate. Doubling the bandwidth of an ADC have a cost. You might argue that highly parallell ADC designs avoids excessive bandwidth demands, but doubling the framerate would still double the bandwidth of each ADC, evrything else being equal.
While your theory may be right, I think it is accurate to claim that recent high-quality DSLRs tend to be either top-performers at low ISO or at high ISO, never both. Who knows if this is a technical thing or a marketing thing.
You make some very good points, and I generally agree. I think one thing you may be leaving out is the advancements in Analog to Digital Conversion over the last four years or so. It is indeed true that as ADC frequency increases, so to does it's contribution to noise. However, the current approach to ADC taken by all manufacturers except Sony is to parallelize buckets, rather than each individual row or column, and process each of those buckets by separate "global" ADCs. Additionally, readout is usually performed on-die in a parallel fashion, however it still is not row or column parallel...usually it is a bucket of columns. Both of those are prime causes for the introduction of banding noise.
Canon currently uses an eight channel per DIGIC 5+ chip approach to ADC. There are two issues with that. First, there are only 8 channels total in most cameras, or 16 channels in the 1D X. That is significantly lower than the 5000+ channels you would have with a true column-parallel ADC approach. Additionally, the ADCs are off die, beyond a bus, on separate chips. That has the consequence of requiring a high speed bus as well as ADCs that operate at a significantly higher frequency than would be required for true CP-ADC. If you put the ADC on-die, you eliminate the bus and the need for high frequency ADCs. You still need a bus for transferring the digital signal out of the sensor, but at that point it is digital, so the signal would not be susceptible to additional noise due to the electronics.
So, yes, a higher frame rate can indeed exacerbate problems with noise. I would agree that the difference between the 7D and the 60D is partly due to the higher frame rate. However, Canon does have a number of patents for parallel readout and image processing on the image sensor die. I am 100% certain they have a form of block bucket readout, which divides the sensor into two dimensional blocks for parallel readout. They also seem to have a form of column-parallel analog readout and amplification with power disconnect, which is apparently capable of nearly eliminating electronic noise that is normally introduced during amplification and read. According to the press releases that announced the 120mp APS-H prototype, they also have on-die image processing. I have not found any patents that seem to describe what that is in detail, but it sounds suspiciously similar to Sony's CP-ADC with Digittal CDS (which is the primary reason Exmor is such a low read noise sensor.)
If we apply current generation technology to the 7D, I it should be more than possible to achieve a high frame rate readout with significantly lower read noise. I cannot say if it could be as low as Exmor (Sony did themselves good with the design of that sensor), but with hyperparallel low frequency on-die ADC and on-die image processing (i.e. noise reduction a la Sony Digital CDS), next-gen Canon sensors could perform at a high frame rate, offer good low ISO performance (maybe not quite as good as Exmor, but certainly a hell of a lot better than any current Canon sensor), good high ISO performance (they already solved that problem, and read noise doesn't play a significant role above around ISO 800), with higher density, higher megapixel designs.