In playing around with extending the allowed transfer time of the bi impulsive transfer, it looks like the simulated annealing code is buggy. It can be replicated just in testing from transfers from equatorial kerbin orbits to minmus. Probably need to see what the actual porkchop plot looks like for those kinds of transfers, and meditate on how the annealing code is getting lost. Might just be simple bugs, but might need to shift to a basin hopping algorithm like what is actually used in practice for this kind of problem.

Did this awhile ago