For starters, with the understanding that different systems will have different random number sequences, we need a simple convergence test, which we can compare with numdiff or some other utility within a tolerance for correctness. Ideally, each individual stochastic simulation algorithm would be tested this way.