As I understand it, streaming tables and materialized views produced by Lakeflow Spark Declarative Pipelines (DLT) are always Unity Catalog managed tables , there's no LOCATION/path option on create_streaming_table or apply_changes.

Is that correct? A few questions:

Is there any supported way to make an SDP output a true external table?

If not, what are the recommended design patterns when you need the data as external — e.g. a designated

managed storage location at the catalog/schema level, a pipeline sink writing to an external path, or a downstream job that re-writes to CREATE TABLE ... LOCATION?

Any reason to prefer one pattern over another for a Gold SCD2 dimension?

Thanks!