The correct choice is Introduce an Amazon SQS queue to accept requests from API Gateway, and have each microservice poll and process messages at its own pace. This option provides a durable buffer and lets the backend consume work at a rate it can handle so bursts do not overwhelm the services.

Decoupling the API from the microservices with SQS smooths demand during a 15 minute promotional spike. SQS stores requests durably and each microservice can poll and process messages asynchronously which prevents sudden overload and allows eventual processing of all requests.

Distribute the microservices across three Availability Zones and use AWS Backup to schedule regular EBS volume snapshots improves fault tolerance and data protection but it does not add elasticity or a buffering mechanism to absorb traffic spikes so it will not prevent overload during short bursts.

Place an Application Load Balancer in front of the microservices and track Amazon CloudWatch metrics for traffic and latency helps distribute traffic and provides observability but it does not queue requests or create back pressure so the backend can still be overwhelmed if it cannot scale quickly.

Configure Amazon API Gateway usage plans with throttling and quotas to limit request rates to the backend can protect downstream systems by limiting request rates but throttling will reject or delay client calls and does not ensure that all requests are eventually processed like a queue does.

Full AWS Practitioner Certification Question