A metric alert describes a condition over a resource signal. It does not become useful merely because the threshold is configured. At Alder Rail, operations created an error condition but never tested whether the intended responder received the notification or knew how to verify recovery.
Keep metric scope, aggregation, evaluation frequency and window separate from the notification path. Logs provide event detail; metrics summarize measurements. Missing data can mean quiet operation or broken observation, so its interpretation must follow the signal. These tasks create local alert-condition records and choose the additional checks a real service needs. The browser does not evaluate live metrics, create action groups, page responders, or run remediation. An alert setting passing here proves only that the requested local condition and target are stored.
Configure the signal and verify the human response path separately.