Thanks for the details. The way it works is once you set it when you first define the job cluster it does not change when you later remove driver_node_type_id from the spec (i.e., omit it), Databricks does not automatically revert it to match the node_type_id.

That’s because the Databricks Jobs API treats unset fields as “leave existing value unchanged”, not “reset to default.” This is by design — the API performs a patch-style update, not a full replace.