Have you worked on a recent idiomatic development setup (dockerised local development, top level imports of everything and plenty of setup at the top level too, people unfamiliar with how to manage .pyc files so they simply disable them...)?

Common libraries like requests or sqlalchemy take 300-500ms to import (eg. try `time python3 -c 'import requests'` and contrast just `time python3 -c ''` which is python startup overhead).

As I said, tests run in sub 10ms, but from issuing pytest to completion it's usually 5-15s.

Ah, I see, so the setup time is very slow. I don't work in python much but I've worked in a few other languages with slow startup, and amortization is your friend. It's hard though when you have a small module with 'only' 300 tests and your test is 6ms of code that works out to 40ms once setup and teardown are included. I haven't had many opportunities to have the "well maybe you should be making bigger modules" conversation but I am ready for that moment to arise.

This is usually the point at which I pull out a 'watch' implementation, since the 5 seconds it's going to take me to switch windows and hit 'up' the right number of times counts too, if we're comparing apples to apples.

That said, one of the last times I had a unit testing mentor, I walked into a project that ran 3800 tests in about 7 seconds, and then started poking around trying to figure out who was materially responsible. (He didn't know much more than me from an implementation standpoint, but boy was he good at selling people on test quality.) If that had been 20 seconds it would have still been lovely, but it wouldn't have grabbed my attention quite as much.

Last time I played with python adding one empty line to my source code was slowing execution my ~8ms.