Test runs used for scoring can sometimes revert to previous browser versions

[DanielRyanSmith]

On September 14th, a test run of Chrome ran with labels "master" and "stable" for Chrome version 116. This came after a Chrome 117 was used for a run with the same labels on September 12th.

https://wpt.fyi/results/?diff&filter=ADC&run_id=5096725200502784&run_id=5118043404894208

This instance of running Chrome 117 as stable and then reverting to 116 resulted in a jump in the Chrome interop score for stable, then reverted that score.

This is likely a transient issue, as subsequent runs of stable have so far been using Chrome 117. If this becomes problematic in further scoring, or additional runs continue to fluctuate between using different browser versions, further investigation and action may be necessary.

cc @nairnandu

[gsnedders]

Annoyingly, https://github.com/GoogleChromeLabs/chrome-for-testing/blob/gh-pages/known-good-versions-with-downloads.json doesn't have any history (presumably force-pushed each time). But given we simply download what's apparently a stable version, this is presumably a problem on the Chrome infra side?