The purpose of the examples would not be to keep those out of tracking whom you'd expect to deactivate (because you are right, it becomes entirely their respons

The purpose of the examples would not be to keep those out of tracking whom you'd expect to deactivate (because you are right, it becomes entirely their responsibility once they are actively presented the option), it would be to encourage trust with those who don't really need to be that level of paranoid (but who often are). Much better than the usual "trust us, everything is so incredibly safe that we don't even know why we have the option to deactivate". Most users will take the "incredibly" at face value and not believe it, and they would be right because nothing is ever that safe. But a risk model like "theoretically there could be a bug in that totally innocent feature use counter that accidentally included a random heap sample in the upload that could include sensitive data" would be acceptable to many. Could a liar smooth-talk you with those examples into pressing accept? Sure. But that bad actor could also ignore the setting or not show an option at all.