are "test_one_each_config" kind of autotest/generated test useful?
Right now we have those automatically generated tests often called "test_one_each_config" which in theory seems like a great idea. But, we've encountered several issues with them:
- they are super slow (more than 5min on py3-web)
- we don't understand them
- when they bug it's generally a nightmare to understand why and how and debugging is very hard
- I don't think anyone has been writing them anymore (everyone seems lost and scared when I mention them)
How could we improve this situation? Should we remove them instead or are they useful?