this is a saner default for large answers. instead of waiting for
probably lost/thrown away fragmented packet, do the query over TCP
instead