It seem fairly logical to me? If a config change causes services to crash then rollout stops … at least in every phased rollout system i’ve ever built…