The correct answer is B because the company needs order processing to begin immediately when new orders arrive and also needs a history of all order processing attempts in the most cost-effective way. Amazon EventBridge is a serverless event bus that can respond to events as they occur, which makes it a strong fit for near real-time order processing. By using an EventBridge event pattern and a rule that invokes AWS Lambda when a new order event is received, the solution avoids polling and processes orders only when there is actual work to perform.

This event-driven design is more cost-effective than running scheduled checks or maintaining EC2 instances. AWS Lambda provides fully managed compute with automatic scaling and charges based on actual execution time. Amazon DynamoDB is a highly scalable and fully managed NoSQL database that is well suited for storing order-processing results and maintaining a durable history of processing attempts with low operational overhead.

Option A is incorrect because invoking Lambda every minute to check for new orders is a polling model, which introduces delay and unnecessary invocations. Option C is incorrect because using an EC2 instance to poll for events creates higher operational cost and overhead. Option D is less appropriate because Amazon EFS is not the best storage choice for durable event history or structured tracking of order-processing attempts.

AWS architectural guidance favors event-driven, serverless patterns for asynchronous order processing when immediate response, scalability, and low cost are required. Therefore, EventBridge + Lambda + DynamoDB is the best solution.