What they really need to do is to simulate data for permissions that are rejected.
For example, if I reject location permissions, then play back a random GPS trail in a randomly selected city on the planet, complete with simulated error and drift. If I reject Wi-Fi scanning, then show a constantly changing set of fake access points. If I reject camera, then play back some cartoons or deepfaked video as a camera device.
The app should never have to know its permission request was denied.
There's two ways Google could solve that: either make better fake data that isn't trivially detectable, or make a rule that trying to detect that gets your app banned from the store.
That would be a fun challenge: given access to all sensors except for the camera, write an app that creates fake camera data.
If Google worked on that, results could be fairly creepy for those who store their data in their cloud. They can probably infer what you look like from your photos (you’re the one appearing in selfies most), know what weather it is locally, know that you’re, for example, looking at the Eiffel Tower, and maybe even have a photo from 2 minutes ago made by another user from the same place.
I think Google certainly could make a fake address book that’s creepy for many users by looking at the address books they have.
The app should never have to know its permission request was denied.