A developer creates an Amazon API Gateway REST API that has a usage plan.

The correct answer is D because the error is occurring when the Lambda function calls the third-party fulfillment service , and the log message clearly states “Rate limit exceeded from fulfillment service.” This indicates the downstream system is throttling requests. In AWS best practices for integrating with external services, developers should use retries with exponential backoff to handle transient throttling and service rate-limit responses gracefully.

API Gateway showing 5XX errors instead of 4XX errors also makes sense in this design. The client request reaches API Gateway successfully, and API Gateway invokes Lambda. The failure happens inside the backend processing path when Lambda calls the external fulfillment API. Because the backend integration fails, API Gateway surfaces a server-side error rather than a client-side error.

Option A is incorrect because increasing API Gateway throttling limits would allow even more requests to reach Lambda, which could worsen the rate-limit issue against the external fulfillment service. Option B is incorrect because provisioned concurrency helps with cold starts and startup latency, not downstream rate limiting. Option C is incorrect because increasing Lambda memory might improve CPU performance, but it does not solve an external API returning rate-limit responses.

By implementing exponential backoff and retry logic , the Lambda function can reduce the frequency of immediate repeated requests to the third-party service, give the service time to recover, and improve overall request success. This approach aligns with AWS guidance for fault tolerance and resilient service integration patterns when working with throttled downstream dependencies.