NSTD.6:5.5 - Evaluation-to-improvement repair input without process theatre

NSTD.6 does not create a big improvement program. It creates result rows. A repeated improvement loop needs only:

NarrativeRenderingImprovementLoopInput@Context:
  objectVersionRef: narrative rendering version, including a candidate
  evaluationResultRefs: selected `NSTD.6` rows
  improvementAim: raise one declared value without lowering protected trade-offs
  allowedChangedSlice: wording, source-return link, ordering marker, viewpoint repair, engagement device, generated source plan, or learning task
  protectedTradeoffSet: source fidelity, owner routing, engagement, cost, reader burden
  expectedReEvaluationForm: rerun the affected `NSTD.6` rows and any neighbor owner checks

If the change is “regenerate until better”, the object version and changed slice are gone. If the change is “add a source-return link and an analogy-stop task”, E.23 can operate and NSTD.6 can re-evaluate.