I work with a lot of large datasets. This seems like it would degrade the performance of the system greatly. Plus, if you have a layer with a feature count over your feature threshold in your service, you would never see the other features when zoomed into other areas. It would be nice if the service was smart enough to provide you a filtered extent upon layer creation and updates to the extent if the service has changes commited to it. I really do appreciate the suggestions.