If more than one connector is available to service a user, and all those connectors are labeled as primary connectors, connectors will service the users according to a "round robin" principle. Say there are three connectors associated with the same group of users, the system will consider the load on all three connectors and which connector last serviced the user and, all other factors being equal, will select the next server in line to service the user.
This architecture can also be set up to include a failover scenario where a "failover connector" is dedicated to take over for several or all connectors in the occasion that they fail or are taken out of service for any reason. The failover connector is not used until one or more of the connectors are overloaded. Once this happens, the failover connector assumes the entire traffic burden, regardless of whether or not there are other connectors available to help share the processing load.