A checkout API running on Kubernetes behind an ALB with Redis and PostgreSQL dependencies starts returning intermittent 503s after a deploy while synthetic checks still pass. The team uses Datadog monitors for p95 latency, 5xx rate, queue age, and a 30-day availability SLO. Which statement best captures the SRE objective for the service?
