Is this a project you guys are going forward with, or just a proposed solution to the problem?
I wonder about the game theory with systems like this. I’m assuming the challenger gets a payout, garnered from the cheating solver, although you didn’t say this. (Otherwise, where will the challengers come from?)
Imagine this system gains widespread use, with thousands or millions of computations served each day. In a mature system, we could imagine that the overwhelming majority of computations will be honest. Because of this, challengers would have to check very many solvers’ work before being “rewarded” by finding a cheater.
In this case, can we actually rely on a cheating solver to be discovered? Can we guarantee that each piece of computation has been checked?