Running Envoy proxy through CRI

** When upgrading Istio sidecar:**
In order to change the sidecar container, the pod needs to be restarted, so the application container will be restarted as well => it will cause downtime for that specific pod.

In order to change the sidecar container without restarting the pod ⇒ We have to run the sidecar out of pod (community has idea that running Envoy through CRI. Unfortunately, there isn’t any formal proposal yet) and Envoy supports hot restart meaning you start a new Envoy binary (could be different version) and traffic is shifted to the new Envoy without traffic loss [1]

Could anyone have any idea on this, please.