After looking over db_object.cc, it looks to me like there's a straightforward path to replacing sqlite3 with lmdb. This may be a "weekend project" or it may turn out to be a more massive endeavor, so let's give it a try. Supporting other data layers in the future is a "stretch goal"