It would be very useful to look at the distribution of FirstContentfulPaint, etc. histograms on different effective connection types. This should first be limited to slow 2G. Understanding paint timing for users that are on slow 2G will give a good comparison for interventions that use Slow 2G as a trigger. Further, we can understand what the page load timing distributions look like on Slow 2G.

Key metrics should include FirstContentfulPaint and ParseStart.