Is there any strategy for dealing with failed objective evaluations by somehow ignoring the failed evaluation? For example, I try to optimize a CFD-based setup with the meshing failing from time to time.