A few questions to help get a better understanding of the situation and may to spark some more conversation:
How many iterations of this update statement are being run as part of the ETL job? Are you updating a primary index field or otherwise indexed column? Do you have RI constraints in place on this table (particularly this column)? Are you seeing blocking (deadlock) issues in Teradata Manager or Viewpoint created by the overlapping intervals as a result of the "poor" performance of the original update?