> A “burn all the packaging down and start over” path?
In Python land one of those seems to come along every couple of years. They start over and implement a solution for some subset of the problem space. Then they realize the Python packaging mess is much, much bigger than anticipated and progress stalls. At that point there are 15 partial "standard" packaging solutions for Python, where there were 14 before. None of them are feature complete, and they don't really coexist well or at all.
In Python land one of those seems to come along every couple of years. They start over and implement a solution for some subset of the problem space. Then they realize the Python packaging mess is much, much bigger than anticipated and progress stalls. At that point there are 15 partial "standard" packaging solutions for Python, where there were 14 before. None of them are feature complete, and they don't really coexist well or at all.
Cue the obligatory https://xkcd.com/927/ and https://xkcd.com/1987/