We used to only cleanup path_get_cache and not path_send_cache causing
issues when a peer disconnects and then reconnects with the same ID.