Aaron Roodman at SLAC points out that we should expect to need to iterate quickly in Commissioning and early Production as we refine algorithms to meet the Science Requirements. He feels these iterations are likely to involve publishing results to the community and obtaining feedback (including code fixes). Trading off coverage in area or time and "certification-level" SDQA in favor of more and faster iterations on unofficial data subsets may be worth it, even to the extent of giving up a DR or two.