As a result, we're not directly measuring the potential performance improvement from changes such as flutter/engine#50521 . From doing some research, this was done intentionally - speciifcally to ignore the warm up frames / shader work done in 2019.

Even if this logic was removed, because the test runner is responsible for connecting to the application and enabling the timeline range, the first frame might not be present and/or this could be flaky.

We may need an alternative test harness to capture first frame times.