when should you start thinking about database connection pooling

hey everyone, ive been seeing a lot of discussions lately about caching, read replicas, and even cdns, which are all great for scaling reads. but im curious about the other side of the coin: managing database connections. weve hit a few snags where application servers are spending noticeable time establishing new connections to the database under heavy load, especially during peak hours. it got me thinking, what are the common indicators that signal its time to implement a database connection pool? are there specific metrics you look at, or is it more of a 'feel it when it hurts' kind of situation? whats your go-to strategy for this