A shared-state client is useful only if it receives events that it can process in parallel. For example, a shared-state client makes sense if it will receive events from one or more publishers and it can process those events in any order (i.e., SHARED_ORDER_NONE). It also makes sense in cases where events must be processed in order by publisher (i.e., SHARED_ORDER_BY_PUBLISHER), and the events come from multiple publishers.

A share-state client is not useful in a situation where it receives events from only a single publisher and those events must be processed in order (i.e., SHARED_ORDER_BY_PUBLISHER). Under these circumstances, the shared-state client would be required to process all events serially. It would provide no performance benefits, because, in this case, the shared-state client would never receive any events that it could process in parallel.