Live Chat: What is the difference between Queue Size Per Agent and Capacity?

Queue Size Per Agent determines how deep the queue can go. For example, given 10 users with a QSPA of 3, up to 30 customer can be "on hold" while the ten agents do what they're doing. This has no effect on how many chat windows agents can have open, only the number of queued customers. Capacity is a measure of how busy users are. This includes chats and other work items. Note that if a user has open tabs, it consumes capacity. This may be what's happening. This may help you:

Omni-Channel capacity relies on the tabs of items that are routed within the console. If you see that agents appear to not have any chats/work items routing to them even when available and capacity appears to be open, have them check their open tabs. If there are any tabs that were routed chats/work items still present in the console, these will consume capacity.

So, if the user is working on other things than just chats, they'll only be able to open a few chats, and if they don't close their completed chats, they'll not get any new chats. You may want to increase their capacity if you think it will help, but you also need to make sure they're closing their work items when they're done with them.

For example, if an agent has a Case routed to them, and a case has a Capacity Usage of 2, with a total Capacity for the agent of 5, then they can only have at most 3 open chats. They can only work so many work items at a time.