Degradation Mode
A defined, intentional reduced state of service that the system enters under specified conditions, with known behaviour and known exit criteria.
Systems degrade whether or not it was designed. The distinction is between degradation that is an enumerated state and degradation that is an emergent accident.
A defined mode specifies four things: the trigger (a dependency circuit opening, load above a threshold, an error budget exhausted), the behaviour (which features are disabled, what fallback data is served, what users are told), the visibility (a metric and an alert, because a system that degrades silently stays degraded), and the exit condition (automatic on recovery, or an explicit human decision).
Common modes worth designing in advance: read-only mode during a primary database failure; cached-only mode, serving stale data with an age limit and telling the user; essential-features mode, disabling expensive optional functionality; queue-and-defer mode, accepting work and processing it later.
Two rules keep them honest. Degraded must be visible — dashboards, alerts and, where it affects users, the interface and status page. And degraded modes must be tested, because a fallback that has never run is untested code on the path you rely on when things are already going wrong.