This allows us to fail fast if the node doesn't exist, and to record node status
changes if we fail to 'allocate' a CIDR.