I often use the third level QVD products from the second level and not the binary load.
In the third level I can have different documents: they read all or some of QVD (the model) created by the second.
I do not think there is an advantage in doing 1, 2, or n levels.
I believe that there is an advantage in choosing the number of levels depending on the situation.
I enclose an old PDF that highlights the pros and cons based on the number of levels