The EJB client testsuite currently requires modelling cluster state "by hand"; in other words, within a test case, when we want to model starting a cluster of nodes, having nodes leave the cluster, or crashing a node in the cluster, the test case writer needs to perform operations on the cluster nodes within the test to make sure that module and topology updates are sent out at the correct time and with the correct content. This is an entirely manual process, is tedious, and needs to be simplified to make it less error prone.