Reconnect to failed configured node not possible, even if the node is up again

In this scenario we have 3 servers running in a cluster (node1, node2 and node3). A standalone client is configured for only a single node of the cluster (e. g. node1).

Now the standalone application is started and is continuously calling EJBs on the cluster. After the first invocation the client get's the cluster view and is distributing the traffic successfully between all 3 nodes.

Now node1 is being stopped. The client continues to distribute the traffic to the 2 outstanding cluster nodes.

After a while node1 is started again. Once the node is up, the client will not consider this node anymore and will not send traffic to this node, even though it's up.