At a $previous_job I basically also did what the post is describing.

The "best" thing we did was actually using a "template database": https://www.postgresql.org/docs/14/manage-ag-templatedbs.htm...

We would start a Postgres Process. We would create a new database, run all of our migrations and basic data bring up. Then we would create a new Database per Test Suite, using the one we just ran migrations as the Template.

This meant the initial bring up was a few seconds, but then each test suite would get a new database in a dozen milliseconds (IIRC).