Mount point is registered with incorrect master on one node

From time to time, clustering suite fails on HTTP 500 [0].

Problem accessing /restconf/config/network-topology:network-topology/topology/topology-netconf/node/netconf-test-device/yang-ext:mount. Reason: <pre> Server Error</pre></p><h3>Caused by:</h3><pre>java.lang.IllegalStateException: Can't create ProxyReadTransaction at org.opendaylight.netconf.topology.singleton.impl.ProxyDOMDataBroker.newReadOnlyTransaction(ProxyDOMDataBroker.java:76) at org.opendaylight.netconf.sal.restconf.impl.BrokerFacade.readConfigurationData(BrokerFacade.java:171)
...
Caused by: akka.pattern.AskTimeoutException: Recipient[Actor[akka://opendaylight-cluster-data/user/akka.tcp:opendaylight-cluster-data@10.29.13.243:2550_netconf-test-device#1278304265]] had already been terminated. Sender[null] sent the message of type "org.opendaylight.netconf.topology.singleton.messages.transactions.NewReadTransactionRequest".

This can probably happen, when device mount point is deleted, but cluster singleton service manages to mark one cluster node as a new master just before mount point is deleted on this node.
Node initiates connection to device. Although it is closed by received delete event, it starts schema setup, which runs in dedicated thread and isn't cancelled. Mount point is registered when setup completes. Mount point without connector hangs in the mount point service and it is impossible to create new mount point with the same id, hence java.lang.IllegalStateException: Mount point already exists int the log.

[0] https://jenkins.opendaylight.org/releng/view/netconf/job/netconf-csit-3node-clustering-only-carbon/513/