Very excited to read this! The question of why mechanistic studies seemed to suggest mask efficacy while mask policies failed always niggled at me - what were we missing in that black box?

At the end of the day, when unit tests pass and end to end tests fail, you realize that you're testing the wrong thing, and you have to go with the actual system behavior. Still, very glad to have a satisfactory answer about why the mechanistic tests results weren't transferrable to the real world setting.

Unrelated, but enjoy this gato treat! https://youtu.be/YiE8V7sqjvw