What processes managed RedShift connections that would explain initializing connections being stuck for 2m when connection count is < 250.

Hello, I'm trying to understand how RedShift manages connections.

I understand that RedShift has a connection limit of 500 which we aren't close to hitting, however we've will intermittently see connections be stuck at initializing for nearly 2m before finally resolving. This is on a new Lambda with normal pg to connect.

Where is connection handling processed, does it have a queue? Does Redshift need to do additional work when using temporary credentials over connecting with username/password?