In serverless world for sure but in old-school architecture it's common to use persistent connections to a database which make connection pooler less essential. Also the last time I did check (many years ago admittedly) connection poolers didn't play well with server-size prepared statements and transactions.