how important was being able to disable postgres full page writes under zfs? that was almost a throw-away line in the article but could be really important for some pg workloads i'd think? certainly got my attention