When defining external tables using formats CSV, JSON, TEXT, BINARY any query on the external tables caches the data and location for performance reasons, so within a given spark session any new files that may have arrived will not be available after the initial query. How can we address this limitation?