CircleCI employs a series of practices to ensure safe continuous delivery of services deployed on Kubernetes, with each service independently contained in its own Git repository. By leveraging Kubernetes' ability to roll out new code pod by pod, CircleCI can maintain multiple versions of code in production while minimizing risk. Strategies include deploying small changes frequently, using readiness probes to detect configuration errors before traffic is routed to new pods, and employing tools like CircleCI, Rollbar, and Slack for deployment visibility and tracking. Additionally, CircleCI uses RabbitMQ for asynchronous communication, ensuring message delivery by delaying acknowledgment until actions are completed, and incorporating unique identifiers to avoid duplicate processing. These practices aim to enhance the reliability and safety of the continuous delivery process.