Why are there many SLP driven duplication jobs for 1 SLP that uses the same retention for the duplication jobs?