This paper may well serve to (give us ideas for how to) make precise the idea of measuring the deviation of solver behavior from the intended behavior. @degregat @AHart I came across this paper when I was wondering how “your” Monad \mathcal{M} in the abstract intent machine could be made slightly less abstract in that we distinguish between the probabilistic nature of the P2P network and the non-deterministic character due to choices of solutions over sets of intents that are active for solving at every given point in time.
Btw: this paper is probably one of the better starting points to delve into this matter.