Understanding round-robin mode on a BIG-IP pool

Hi, hopefully super simple question from someone with very little F5 experience.

We have a pool with five members, configured for round robin. I think our expectation was that each new connection would go to the next server in the pool and things would keep cycling round in that fashion for an even distribution.

What we are seeing is that connections are clumping together on 1 or 2 servers, with several showing 0 connections at any one time. Just wondering what we have missed here and how much setup is required for a simple round-robin environment.

Would the fact this was all originally set up for Observed Node impact things?

Any responses are much appreciated, thank you