A Blue/Green deployment strategy is the most appropriate choice because it directly addresses the core requirements of minimizing downtime and providing instant rollback. This pattern involves running two identical production environments: "Blue" (the current version) and "Green" (the new version). Traffic is switched from Blue to Green only after the Green environment is fully deployed and tested. This switch is nearly instantaneous at the load balancer or service routing level. If a critical failure occurs, rollback is also instantaneous, as traffic is simply routed back to the still-running Blue environment.

B. Canary Deployment: This strategy involves a gradual rollout to a small subset of users. While it minimizes risk, it does not provide an "instant" rollback for the entire application; it requires scaling down the new version.

C. Deploy to a new GKE cluster and manually switch the DNS records: Using DNS for traffic switching is not ideal for minimizing downtime or instant rollback due to DNS propagation delays and client-side caching, which can take minutes or hours.

D. In-place Rolling Update: This is the default Kubernetes strategy where old pods are gradually replaced by new ones. A rollback is not instant; it requires initiating another rolling update to deploy the previous version's configuration.

1. Google Cloud Documentation, "Application deployment and testing strategies": This document explicitly describes the Blue/Green deployment strategy. It states, "This technique can eliminate downtime during a deployment. Also, rollback is instantaneous because you can just switch back to the old environment if something goes wrong." This directly supports the choice of Blue/Green for the stated requirements.

2. Google Cloud Blog, "6 strategies for application deployment on GKE": In the section on Blue-Green deployments, the article lists "Instantaneous rollout/rollback" as a primary advantage. It contrasts this with rolling updates, where a rollback "can take a significant amount of time, depending on the application."

3. Kubernetes Documentation, "Deployments": The official Kubernetes documentation describes the default "RollingUpdate" strategy. It details how pods are terminated and created sequentially, which clarifies why a rollback is a time-consuming process and not instantaneous, making it less suitable than Blue/Green for this specific scenario.