[Java] Properly fix shouldStartClusterThenElectNewLeaderAfterPartition test intermittency, i.e. await election termination before sending more messages into the Cluster. This is to deal with cases of unexpected vote request which leads to multiple rounds of elections being executed.

Author: vyazelenko

aeron-system-tests/src/test/java/io/aeron/cluster/ClusterNetworkPartitionTest.java

@@ -107,14 +107,15 @@ void shouldStartClusterThenElectNewLeaderAfterPartition()
         cluster.sendMessages(50); // will be sent to the old leader
         Tests.await(() -> firstLeader.appendPosition() > initialLeaderLogPosition);
 
-        final TestNode secondLeader = cluster.awaitLeaderWithoutElectionTerminationCheck(firstLeader.index());
-        assertNotEquals(firstLeader.index(), secondLeader.index());
+        final TestNode interimLeader = cluster.awaitLeaderWithoutElectionTerminationCheck(firstLeader.index());
+        assertNotEquals(firstLeader.index(), interimLeader.index());
 
         cluster.awaitNodeState(firstLeader, (n) -> n.electionState() == ElectionState.CANVASS);
 
         IpTables.flushChain(CHAIN_NAME);
 
-        cluster.awaitNodeState(firstLeader, (n) -> n.electionState() == ElectionState.CLOSED);
+        final TestNode finalLeader = cluster.awaitLeader(); // ensure no more elections
+        assertNotEquals(firstLeader.index(), finalLeader.index());
 
         cluster.reconnectClient();
         cluster.sendAndAwaitMessages(100, 200);