[AlexanderKaran]

Brought Variance back as a top-level variable. So every time a full run of a benchmark happens, and we re-create its top level result we add a variance key that compares it to the last version, like so:

{
  "name": "coldBuild",
  "runs": 5,
  "iterations": [
    { "timeMs": 955 },
    { "timeMs": 930 },
    { "timeMs": 925 },
    { "timeMs": 936 },
    { "timeMs": 967 }
  ],
  "timeMs": {
    "mean": 942.6,
    "median": 936,
    "min": 925,
    "max": 967,
    "variance": {
      "comparedToMeasuredAt": "2026-10-27T03:16:33.289Z",
      "meanPercent": -9.10,
      "medianPercent": -1.89,
      "minPercent": 3.24,
      "maxPercent": -29.88
    }
  }
}

Only compares if the versions are the same. Then, when the stats PR is generated, a special CI/CD check runs; if any of the benchmark results are bigger than a 10% swing, it flags it to re-run the test.

We can, of course, lower this number and later configure different values per benchmark. Not storing historical data for now just compare the current one each time until you bump versions

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	framework-tracker	8d31880	Commit Preview URL Branch Preview URL	Oct 27 2026, 12:49 PM