The author writes that this is a contrived case but I was doing Monte Carlo simulations 10 years ago, i.e. I wanted exactly this case.

In the end, I just let them run on one thread and could never work out how they could be 10x slower with threading - until today. Thanks.