I broke the system down into components: API gateway, authentication layer, message queue (Kafka), microservices handling payment processing, retry logic, dead letter queues, and monitoring/logging. I explained trade-offs, consistency models, and why eventual consistency makes sense in this case. I also discussed database options and failover strategies for high availability.