The mongocxx driver API is designed such that you get a connection object from the pool and then you get a database from the connection. Any operations then happen over that connection I assume? So my question is do I need to keep the connection object alive while I use database/collection objects? e.g.

CollectionWrapper::CollectionWrapper {
 connection = pool.acquire();
 db = connection.database("my_database");
 m_collection = db.collection("my_collection);
} // << connection object destroyed here but member variable m_collection is still alive

Or does the fact that I am keeping the collection alive is enough to ensure the connection is not returned to the pool?