From #1626:

Once we have existing functionality shored up, we need to start modernizing. Our model uses local prevalence (as measured by officially reported testing numbers) to adjust risk. The concept of adjusting your activities and safety measures based on risk is a really compelling microCOVID feature. That said, we know the amount of official testing has reduced, as a secular trend independent of the cyclical waves of infections. Personally, I haven’t seen a line in my neighborhood for the testing kiosks for a while now. The milder acute symptoms of Omicron strains, the popularization of home testing and a general weariness of safety measures are surely causes of the trend towards underreporting.

We should find a way to make an adjustment to try to correct for this, and ensure we are expressing to users the limitations of this good faith measure along the way. Again, I lean towards the pragmatic.