I said most cases. What you are describing are the exceptions, not the rule. I don't remember how he phrased it exactly but about a year or two ago Andy Mueller

I said most cases. What you are describing are the exceptions, not the rule. I don't remember how he phrased it exactly but about a year or two ago Andy Mueller tweeted that his biggest wish is that scikit-learn would eventually receive enough support to be able to have the maintainers work full time on scikit. Think about this for a second: I can't recall seeing a single ML project that doesn't rely on scikit to a certain degree and the core maintainers are 8 iirc. That's pretty sad if you ask me.