Andrew Nelson

*Please provide a full list of instructions from an empty docker container (opensuse/tumbleweed), that will allow us to reproduce the problem*.

I think we might close this now. Any adjustments would probably be relaxing of a test tolerance anyway.

Leave for the time being.

Thanks for the feedback @StefanBruens. I appreciate where you're coming from. One difficulty with those constrained tests is that some of them are very slow to run. This is not...

We had a good discussion on this exact topic, I can't remember if it was on the ML or an issue though. @rgommers might be able to remember. I had...

> having this variability in the long run is a good thing Do you mean stability/reproducibility/invariance in the long run?

> after we make this change, there is no simple run-to-run stability by setting a single global state I'm fully behind removing usages of the global `RandomState`. However, it wasn't...