The data does not speak at all. You are drawing inferences based on the data, inferences of a type that are pretty easy to poke holes in by looking at the pretty large class of things rarely used but situationaly useful (often life saving).
This type of bizarre logic makes me feel like I have to habitually use features even when they are not useful in order to protect them from UX designers in case I'm subject to some misguided A/B test.
What data are you referencing? I'm drawing conclusions based on data I have but I don't think you work for the same company as me so I'm not sure how you can speak to confidently about how wrong I am.
This type of bizarre logic makes me feel like I have to habitually use features even when they are not useful in order to protect them from UX designers in case I'm subject to some misguided A/B test.