Porting nontrivial codebase from 'F1 all the way through 'H7 is IMO a hard and bumpy road with potentially major debugging ahead. It of course depends on the nature of the code, the ratio between "true controller" and "general purpose" code.
Now, we are not insiders here, but as others pointed out, it's unlikely ST will add a Cortex-M4 with both ETH/LCD and "controller" stuff like CAN is. The 'F4 came around when the Cortex-M7 was not ready yet. Note that their newest addition, 'G4, used up the "4" number, and is straighforwardly targeted at the 'F3 area of applications (i.e. motor). ETH goes together with LCD more into the "general computing" and "needs more horsepower" area, that's why they pair up better with the Cortex-M7.
So, in short, I wouldn't hold my breath. If you don't want to go the bumpy road, you may try the multi-chip way (either external CAN controller, or cooperating MCUs, one with CAN other with ETH) - although that's not obstacle-less and not price-optimal either, I know.
JW
