Computational cost for proof generation

Here’s a link to the latest flypool experience with z-addr join-split (it’s worse than I recalled):