If a node is joining just as the last member of a cache is leaving, the ClusterCacheStatus on the coordinator may be updated incorrectly. The cache topology is reset to include only the joiner, but the joiner ignores it because the new topology id is set to 0.