> Would it make sense to parallelize it now
> (breaking the consistency with previous distort()
> results) *before* anyone has a bunch of
> differently-seeded distorted meshes depending
> on that consistency?
In case you're asking me: I don't feel like I can answer that question. I guess I won't be using ParallelMeshes anytime soon, so any implementation that lets one influence the randomness in some way will be fine with me.