Oh, I see. You're counting on the GFI not noticing a voltage difference in the returns.
I don't see why that wouldn't work right offhand, but the whole thing is already a Rube Goldberg machine even before you start down that road...
Oh, and the current between the two hots is not static. Your "compensator" would need to vary the return current to keep the two GFIs happy.
EDIT: Oh you said that already. Yeah.
I have my doubts on the 'quick enough' part.
A typical GFI is just a CT coil with the hot and neutral running through it. Any time the current on the CT exceeds the trip current divided by the Te of the coil, the GFI will trip. I can't envision a reaction time fast enough to overcome that arrangement.