During a bulk load we compare each key inserted against the last key inserted so that we don't let the user create a corrupted table; we wouldn't have to do that comparison when called from the LSM merge code because we know it's inserting in the right order.