Consider a CLUSRCVR channel for a minute. You define one of these and then all members of the cluster use the same one to talk to the local QMgr. Similarly, a single SVRCONN channel definition can server connections from many users (think SYSTEM.ADMIN.SVRCONN for instance, with 50 users attached to it.)
The same principle works for RCVR or RQSTR channels. For example, suppose that you create a channel on your local QMgr called BRANCH.TO.CENTRAL with the intention that each branch office uses the channel to send daily sales numbers to the home office. Each of the queue managers can have only one instance of a BRANCH.TO.CENTRAL channel defined. The branch queue managers can have only one instance of the channel running but the central office can have many instances of the channel active at the same time. Typically you would not do this and would either use a cluster or named channel pairs, but it does serve to illustrate the case.
So the rule of thumb is that inbound channels (SVRCONN, RCVR, RQSTR and CLUSRCVR) can have many running instances using a single channel definition while outbound channels can have only one running instance.
Does that help or just muddy things up?