Has anyone practical experience with this scenario? Given an increase in load and concurrency for a service, will the sidecars scale as well? Can the limits be defined?