Decides what a change must emit so a production failure is detected, diagnosed, and verified, and keeps the resulting signal set honest as it ages. Use when a risk survives verification into production, when choosing between a log, a metric, a trace, an event, or a probe, when defining a service level indicator or the threshold an alert fires on, when an incident showed the failure was invisible or the page was unactionable, or when auditing alerts, dashboards, and telemetry for noise, cost, and signals nobody consumes. Do not use to design a pre-production check, place a check in the pipeline, gate a release, or link outcome evidence to a requirement; hand those to test-…
