Before going through something like delayed replication, you really want to consider using zfs or lvm and taking regular snapshots on your hot or warm standby. In the event of the accidental table drop, you can just roll back to the snapshot prior and then do PITR from there.
> Well, yeah. The point was that you possibly could run it for a while to > "catch > up" without taking a new base backup if you desired. You should also keep > copies of it for PITR.
