I did not take notes I could quote but I recall that when they first started offering their new speed test that they had a little disclaimer attached. Something about it being inaccurate because of a lack of data at that point. I would expect that they should have a little more data by now. I don't expect much from GSC info and tools and they don't disappoint in that respect.

I don't expect much from GSC info and tools and they don't disappoint

Words To Live By.

I think the GSC version is inaccurate, and would take it with a very generous pinch of salt.

For all my pages, on both mobile and desktop, GSC reports that performance is "Moderate". The reported "fault" (preventing a result of "Fast") in all cases, is "FCP issue: longer than 1s". This has applied on every refresh since 19 September.

Test any page individually ([developers.google.com ]), however, and the "fault" disappears: all my pages usually score 100 on both mobile and desktop, although occasional glitches occur there too..

Any one of a number of things could cause this, but my pages load very quickly on an actual desktop or mobile, and that is the test a user applies. To me, therefore, the variation in Google results casts doubt on the test, not on my pages.

Several things that will improve Google test results - and which Google recommends - are in my view poor practice (deferring loading of or inlining CSS. is right at the top of that list), but if you want better test results rather than pages that are optimised for your users you can always work through and apply all the recommendations they give.

What has caused the "promotion" of your pages (changing the location of the test server would be one of my first suspicions) will depend on which problem GSC thinks has been resolved, but whatever it is lies on their side of the counter. If your pages perform well in real-world conditions I wouldn't worry about what GSC says.

Much of this reporting is placebo not reality.