The mirror service exists for the purpose of migration, where all blobs exist in the primary subservice and a subset of blobs exist in the secondary subservice. Since the primary subservice is the source of truth until a migration is completed, operations like existence checks need not be performed against the secondary subservices.