- if resource.error bad revision is recieved we http to get a new one - in HA world this can still result in a watch with resource.error bad revision (given a stale replica handles watch) - to avoid a tight loop, back-off the http request