Since in Airflow we now use Hive as a source of truth for dataset availability. we need to update the refinery scripts to not depend on _SUCCESS files but trust Hive (they can generate some though).
Scripts depending on _SUCCESS flags:
- refinery/bin/refinery-drop-mediawiki-snapshots - https://gerrit.wikimedia.org/r/c/analytics/refinery/+/786448
- refinery/bin/refinery-dump-status-webrequest-partitions - checks that latest partitions of webrequest and pageview are processed - could be modernized