Interesting, we've just been looking at something similar at work. Unfortunately we've got a bunch of application config and test data in the database and no fast process for seeding it. We're now building a Postgres Docker image with a baked-in database. There were some tricks for making a small database and it's important to change `PGDATA` to something else, otherwise the `VOLUME` directive (in the official Postgres image) causes a full copy of all the files. It looks promising, though (accepting connections in ~200ms).