This is a bug in the REST API's delete MUC Room service, when Hazelcast clustering is enabled, and there is more than one Openfire/Hazelcast node in the cluster.
The failure happens when a MUCRoom is deleted via REST. It will be deleted on node which received, but it will still exist on the other nodes of the openfire cluster.
In the case of deleting a room, with multi-node clustering enabled, sending RoomRemovedEvent to the other cluster nodes will resolve the issue.
Pull request was merged. I'm going to resolve this issue