If it is indeed true that the 40+mp sensor was rejected for noise, would it not make sense for Canon to consider extrapolation of a tried and tested density? If you consider that at 4896x3264 on a 519 square millimetre chip the 1D Mk4 has a density of 30794 per square millimetre, which is significantly higher than the 24336 on the 5D2 and 1Ds Mk3. Then there is an opportunity to use this wafer at 36 x 24 to provide a 26.6mp FF sensor. As this would have 1.66 x the pixels of the 1D Mk4, then the existing dual Digic IV providing 10FPS could even using current algorithm manage to process 6 FPS.
A 26.6 mp 1Ds Mk4 with around 6318 x 4212 and 6fps might not satisfy the mega-pixel hungry wishing to beat Nikon into submission, but it would provide a sensible step up using a combination of available technology without having to push the price point through the roof. It would also leave the 'blue sky' development bods to work on the 46mp for the 1Ds Mk5 etc. I might be way off, but with 30 years of running R&D it would certainly be a step I would consider.