I can't imagine it being that difficult to verify correctness of the changes -- just compare the results of the original test suite, and the results of the para

I can't imagine it being that difficult to verify correctness of the changes -- just compare the results of the original test suite, and the results of the parallel one -- and try this repeatedly until you're confident you've avoided any missed specifications If the parallelism actually speeds things up, repeated runs shouldn't be too painful :-)