when does connection pooling become a bottleneck

hey everyone, ive been thinking about connection pooling lately, especially in high-throughput applications. were using it a lot to manage database connections, and its generally been a win. but im starting to wonder if theres a point where the pooling mechanism itself can become a performance bottleneck. are there specific metrics or scenarios where youve seen connection pools cause issues? what strategies do you employ to monitor and optimize them, or even when do you consider alternatives like serverless databases that handle connections differently?