the C++ driver should have an option to asynchronously fire off a getMore mesage before the client consumes the current batch of data, and then wait for the response when the client asks for more data than has so far been received.

this would improve replication and cloning performance as client and server can then work concurrently – replication code (repl.cpp) should use this flag.

some care must be taken that the asynchronous more result is waited for if the client decides not to consume all the data from the query (as this would be an option, perhaps one must consume it all and this is simply an assert check that this is done)

the same pipelining feature would make sense in other drivers.