This was found when testing a fix [0] to NETCONF-454 on Sandbox.

During rejoin, singleton service has notified the old master it lost the ownership. Slave mount point was created, registering (code point [1]) for operational changes. That was done when the old master member has topology-operational shard still in Candidate state (while new master has already had updated the operational data in the majority partition).
But the initial change notification has never been received (code point [2]).

[0] https://git.opendaylight.org/gerrit/#/c/64706/2
[1] https://git.opendaylight.org/gerrit/gitweb?p=netconf.git;a=blob;f=netconf/netconf-topology-singleton/src/main/java/org/opendaylight/netconf/topology/singleton/impl/NetconfNodeManager.java;h=e78e9ee9c7ba55984e2b9fbf962861c8a2418aab;hb=220a78adafcfc1280fb2af6970292a749353defb#l129
[2] https://git.opendaylight.org/gerrit/gitweb?p=netconf.git;a=blob;f=netconf/netconf-topology-singleton/src/main/java/org/opendaylight/netconf/topology/singleton/impl/NetconfNodeManager.java;h=e78e9ee9c7ba55984e2b9fbf962861c8a2418aab;hb=220a78adafcfc1280fb2af6970292a749353defb#l73