When I changed the view port size for our tests in T198394 it increased the First Visual Change for Firefox. That's fine but it also introduced what looks like more unstable metrics. I'm gonna go through the graphs and have a closer look, but until then I will try to use the Firefox Window recorder to see if it could be FFMPEG with the current instance size that could have increased instability. That fix comes in https://gerrit.wikimedia.org/r/c/performance/synthetic-monitoring-tests/+/636461