I've been using MongoDB for quite a long time and didn't know about db.requestStart() until this past week. If you're going to call getLastError than it seems like it would be nice to ensure that db.requestStart() has been called. This could be an optional feature to not break existing code.

Or an even better question is why is getLastError a separate command? That has always seemed very odd to me. Why not just add an option where the original request can return the error in the response? Then you would save a network round-trip.