Concurrency problem on startup in PostgreSQL cluster mode

On startup there is a cleaning procedure to remove orphan invalidation. This cleaning is done on each node even on instance with a noDDL repository option. This cleaning should not be done concurrently because it hangs:

DELETE FROM CLUSTER_INVALS
  USING CLUSTER_INVALS i LEFT JOIN CLUSTER_NODES n ON i.NODEID=n.NODEID
  WHERE n.NODEID IS NULL