With regards to dynamic range, the main issue is that Sony's sensor architecture results in dramatically lower read noise compared to Canon. But read noise is not the only factor affecting sensor noise; high ISO noise is the other. Once high ISO noise becomes the dominating factor in sensor noise, Canon's sensors are competitive with Sony's.
The two ways for Canon to fix this problem is for them to redesign their sensor architecture to reduce read noise, or just to jack up the MP count so much that ISO 100 looks like ISO 3200 on a per pixel basis. However, this means that on a print basis, you can't tell the difference between Canon and Sony.
The current EV difference between Canon and Sony is 2.59 EV at ISO 100. Increasing pixels to a factor of 6 gives you about 132 MP, and should completely bridge the gap between Canon and Sony at ISO 100 while making it so that Canon would have better performance at ISO 200.
An increase in MP by 5 times, though, would require 6 times the processing capability. To match the D800's 4 frames per second, you'd need 4 DIGIC V+ processors. This would also likely cut battery life by 3/4ths, and would require a substantially larger battery.
A better way to handle this would probably be to go with a test APS-C camera. Large pixel cameras require humongous processing capabilities on the desktop computer, and with APS-Cs, you could use about the same pixel density with a reduced MP count. This would be about a 51 MP APS-C camera.