I’m wondering if they’ve considered less radical solutions? Something like adding a new thread-based front end to Postgres itself while keeping the processes, and then gradually fixing whatever makes that less than seamless.