not quite (because of how the internal subproblem dispatching works), but yes, a good chunk of the subproblem results should be reusable