control_service_queue_mem_limit default is too low for large clusters

We've seen some cases where this queue overflows because of status reports, particularly with higher mt_dop values or large clusters. We should probably set it to a higher value by default.