The other problem is the loss rate


How do you find and eliminate parts per million (billion?) packet loss across multiple providers with extremely diverse infrastructures, all while under heavy load?

Current tools and measurement infrastructures are working at the wrong scale by several orders of magnitude.