I am skeptical to its architecture. This is 2017 and there are compelling reasons for endpoints to do resolving on their own. To get rid of open resolvers would be nice.

"But .. caching", well, the performance of resolvers is perhaps not as clear cut as that. Performance would suffer for some, but they can clearly handle it, and does it really matter?

Some real world testing would surely be beneficial. In the mean time, I'm not sure about solutions to the resolver data leak problem. It is a solution to a problem we should not have.