When defining external tables using formats like CSV, JSON, TEXT, or BINARY, any query on these tables caches the data and its location for performance reasons. This means that within a given Spark session, any new files that arrive after the initial query will not be available. How can we overcome this limitation?