It was noted in the quarterly review that forward looking speed enhancements would do well to try to measure in terms of median page size load time on varying connection speeds.
So being able to express stuff that way will be good. Realistically, having a distribution curve for different page sizes on different connection speeds can be even more telling.
Having a sense of the constituent parts of the page, particularly ones we want to remove / lazily load, will be helpful. This implies some kind of work on Event Logging schema(s).
Can we do this with existing performance infrastructure or will we need to build something new?