In order to prevent unnecessary alerts and distractions from unstable metrics, Datadog has introduced recovery thresholds. These thresholds work by setting a different value for when an alert triggers and when it resolves, using the principle of hysteresis. When creating a monitor via the UI or API, users can add recovery thresholds to their alert conditions, which apply to threshold alerts, change alerts, and anomaly detection. By doing so, users can increase confidence that an issue has truly been resolved once an alert recovers.