You want to let your team deploy a new Cloud Run revision to production while limiting customer impact and following Google best practices. What should you do?

Gradually rolling out a new revision and splitting traffic is the recommended best practice for deploying changes to Cloud Run. This approach, often called a canary deployment, allows you to direct a small percentage of traffic to the new revision first. You can then monitor its performance and stability with real user traffic. If issues arise, you can quickly roll back by directing all traffic to the stable, older revision, minimizing customer impact. Sending all traffic to the new revision immediately risks a full outage if there's a problem. Asking customers to retry requests shifts the burden of failure handling to them, which is poor user experience. Deploying a second Cloud Run service and asking customers to switch is an inefficient and disruptive way to manage updates, as it requires manual client-side changes and doesn't leverage Cloud Run's built-in revision management capabilities.