DROP NODE leaves data with the old node ID behind in the log tables. It would be possible after rebuilding the node with the same ID that old log rows, still existing on a forwarder, get picked up by a cascaded subscriber.