I have no insider knowledge, but I'm going to guess that they even crawl using actual Chrome browsers once in a while... and penalize deviations from what you serve to the GoogleBot... and also factor in response times for either.

I don't think any single person can predict/know what's actually going on at this point.

But this mitigation would still fail. They might test for different content, but they probably don't include a periodic test like this in their timing tests, and if they did, it would just be an outlier.

In google search console, you can see lighthouse data reported from actual google chrome instances of real users. (I can’t verify this right now)