A payment processing application runs on AWS Lambda in private subnets across multiple Availability Zones. The application uses multiple Lambda functions and processes millions of transactions daily. The architecture must prevent duplicate payments. Which solution meets this requirement?

The correct answer is using an Amazon SQS FIFO (First-In, First-Out) queue. FIFO queues guarantee message ordering and support exactly-once processing, which is crucial for preventing duplicate payments. When a message is successfully processed, it can be deleted from the queue, ensuring it's not processed again. The other options are less suitable: Publishing to an Amazon S3 bucket with event notifications doesn't guarantee exactly-once processing or order, potentially leading to duplicates. A standard Amazon SQS queue doesn't guarantee message ordering or exactly-once processing, making duplicate payments possible. Storing in an Amazon DynamoDB table with streams also doesn't inherently prevent duplicate processing without additional complex logic to track processed items, and streams don't guarantee exactly-once delivery for downstream consumers in the same way SQS FIFO does.