I thought about this all day. Given the scope of the change you made I think no option looks ideal. If we revert the entire feature, it's a big bummer for everybody involved and it might be an overreaction. If we keep the feature at the state of Beta 1, it might become a maintenance nightmare for third-party users. If we agree to a the big change you now see necessary, we are reintroducing the same risk that I highlighted in previous comments.

Please note that this is in no way intended to make you feel terrible. Don't beat yourself up about it. Now, think as the core developer you are: if you are 100% sure the change is worth pushing to 3.9, the best course of action at this point seems to be: find a champion within core who will work with you to quickly reassess the situation and give us an action plan as a team. 

Say, Barry seems like the perfect candidate. If such champion shares your opinion about it being important for 3.9, and is willing to work with you on finishing what you need changing (incl. review), *and* to ensure the existing users of importlib.resources don't see any breakage not documented as of Beta 1, then we can talk about an exceptional late inclusion in 3.9.