Progressive delivery is a deployment strategy that involves gradually shifting production traffic to new application versions without downtime, with Argo Rollouts serving as a Kubernetes controller to facilitate such deployments through methods like blue/green and canary releases. While straightforward for stateless applications, using Argo Rollouts with stateful services like databases or queues requires additional configuration to prevent preview versions from inadvertently accessing production resources. A solution involves leveraging the Kubernetes Downward API to pass ephemeral labels to applications, allowing them to distinguish between "preview" and "stable" modes and directing them to appropriate resources. This requires the application to automatically reload its configuration when labels change, a capability that must be implemented by developers using suitable libraries for their programming language. An example using RabbitMQ demonstrates how this setup can prevent preview applications from accessing production queues, instead directing them to separate, designated queues. The process also includes promoting new versions to stable status by updating labels, with applications automatically adjusting their configurations to use production resources, showcasing a sophisticated approach to managing progressive deployments in Kubernetes environments.