You must design an event-processing solution that scales automatically with the number of incoming events and, on processing failure, moves the problematic event to a separate queue for later review. Which architecture meets these requirements?

This architecture effectively meets the requirements. Amazon SNS provides a highly scalable publish/subscribe mechanism for event distribution. AWS Lambda automatically scales with the number of incoming events from SNS, eliminating the need for manual scaling configuration. The Lambda function's on-failure destination feature directly supports moving problematic events to a separate Amazon SQS queue (acting as a dead-letter queue) for later review, fulfilling the failure handling requirement. The other options have drawbacks: Using EC2 instances with SQS requires managing EC2 scaling and application-level DLQ logic, which is less serverless and more complex than Lambda's built-in features. DynamoDB streams are primarily for capturing changes to a DynamoDB table, not for general event processing, and don't inherently provide the same scaling or failure handling for incoming events as SNS/Lambda. EventBridge with EC2 instances introduces EC2 management overhead and requires configuring the EventBridge target and DLQ, which is less streamlined for simple event processing than SNS/Lambda.