Introduced on day 7
Idempotent pipeline
A data job that produces the same result whether it runs once or several times for the same input range.
In detail
Achieved by overwriting a partition (delete then insert for the date range) or merging on a key instead of appending. Non-idempotent retries create duplicate keys.