How do I achieve zero-downtime deployments for Argo Workflows?

# Example of zero-downtime deployment strategy for Argo Workflows
apiVersion: argoproj.io/v1alpha1
kind: Workflow
metadata:
  generateName: zero-downtime-example-
spec:
  entrypoint: main
  templates:
  - name: main
    steps:
    - - name: deploy-new-version
        template: deploy
    - - name: upgrade-old-version
        template: downgrade
        when: "{{steps.deploy-new-version.outputs.status}} == Succeeded"

  - name: deploy
    container:
      image: myapp:latest
      command: ["/bin/sh", "-c"]
      args: ["echo Deploying new version"]

  - name: downgrade
    container:
      image: myapp:previous
      command: ["/bin/sh", "-c"]
      args: ["echo Downgrading to previous version if necessary"]