Thanks for the quick reply. This is super-helpful! Great to see that I am not overlooking something trivial here. Also thanks for the mcbackend reference, I will try it out also.
As for the assign intermediate expressions part - very good point, thank you. I think I will replace most of my pm.Deterministic variables with normal expressions then.
One small inconvenience with removing the mu = pm.Deterministic(...) is that I was relying on mu.shape to inform the “shape of the likelihood”…