The incremental job does not create any records it just merges updates with the existing dwh records. So when a duplicate key occurs, the incremental will repeatedly and consistently fail until the full load data warehouse job is run.
select period_KEY, period_KEY, COUNT(*)
group by period_KEY,period_KEY
having COUNT(*) > 1