The pattern of falling back to a reduced experience when a dependency fails, instead of failing the whole request.
Graceful degradation is the practice of designing a service so that when a downstream dependency fails, the service returns a reduced but still useful response instead of a 500. Recommendations service down? Show static popular items. Personalization service down? Show the global feed. Recently-viewed cache empty? Show recent purchases instead. Each degradation is a deliberate engineering decision, made before the incident, that turns a hard failure into a soft one.
Most outages are not single-component failures, they are cascades where one downstream's slowness saturates an upstream's connection pool, which fails its upstream, which fails its upstream. Graceful degradation breaks the cascade at the boundary by making each layer survivable when its dependency is down. The technique is the difference between 'one slow database = total outage' and 'one slow database = one feature degraded'.
See the part of the platform that handles graceful degradation in production.