A payments microservice on AKS calls an external provider that intermittently returns 5xx and timeouts under load. Spikes cause thread pool exhaustion and cascading failures in other services that share the same node pool. You must prevent cascading failures and keep unrelated services responsive when the provider degrades. Which two resiliency patterns should you implement?

A circuit breaker is crucial because it prevents the payments microservice from continuously retrying a failing external provider. By opening (short-circuiting) after repeated failures, it allows the provider to recover and prevents resource exhaustion within the payments service, thus stopping cascading failures. A bulkhead pattern isolates the payments service's resources (like thread pools and connections) from other services on the same node pool. This ensures that even if the payments service experiences high load or failures due to the external provider, its resource consumption is capped, preventing it from starving other services and causing cascading failures across the node. Increasing retries with no jitter would worsen the problem by adding more load to an already struggling provider. Fan-out/fan-in is for orchestration, not direct resiliency against external service degradation. Removing timeouts would exacerbate the issue by holding resources indefinitely.