Nodes not part of cluster have marked the Gossip as seen, see #3031

* Problem may occur when joining member with same hostname:port again,
  after downing.
* Reproduced with StressSpec exerciseJoinRemove with fixed port that
  joins and shutdown several times.
* Real solution for this will be covered by ticket #2788 by adding
  uid to member identifier, but as first step we need to support
  this scenario with current design.
* Use unique node identifier for vector clock to avoid mixup of
  old and new member instance.
* Support transition from Down to Joining in Gossip merge
* Don't gossip to unknown or unreachable members.

akka-cluster/src/main/scala/akka/cluster/ClusterReadView.scala

@@ -58,7 +58,8 @@ private[akka] class ClusterReadView(cluster: Cluster) extends Closeable {
             state = state.copy(members = state.members - member, unreachable = state.unreachable - member + member)
           case event: MemberEvent ⇒
             // replace current member with new member (might have different status, only address is used in equals)
-            state = state.copy(members = state.members - event.member + event.member)
+            state = state.copy(members = state.members - event.member + event.member,
+              unreachable = state.unreachable - event.member)
           case LeaderChanged(leader)                          ⇒ state = state.copy(leader = leader)
           case s: CurrentClusterState                         ⇒ state = s
           case CurrentInternalStats(stats)                    ⇒ _latestStats = stats