Very good and interesting article!
A positive deviation from Postgres is that Aurora DSQL does not require connection pooling! Your short-lived connections are handled by the database itself, and you don't have to worry about managing a separate connection pool in front of the database.
But that does still mean we rely on connections right? If something scales (e.g. Lambda) we do have a max connection limit (I think 1k) and then other connections will be refused?