I’d love to see, based on the level of non-determinism perfomance on the benchmark how many times you need to run the benchmark for the change to be relevant (o

I’d love to see, based on the level of non-determinism perfomance on the benchmark how many times you need to run the benchmark for the change to be relevant (or statistically significant if you want). That would be a nice paper.