in master-slave, when there's a network failure or other issue we may play a sequence of operations multiple times - the data will be eventually consistent, but not a snapshot from the master

may want to address this as we move toward repl sets