So I'd say there is some merit in orchestrating this with something else like Airflow or Argo just because it will be easier to debug isolated issues.... But after_pipeline_run feels like the safest place to do this