Just anecdotally, I get the feeling telemetry often does more harm than good, because it's too easy to misinterpret or lie with statistics. There needs to be proper statistical methodology and biases need to be considered, but this doesn't always happen. Maybe a contrived example, but someone wants to show high impact on their next performance review? Implement the new feature in such a way that everyone easily misclicks it, then show the extremely high engagement as demonstration that their work is a huge success. For Git, I'm not sure it would be widely adopted today if the development process was mainly telemetry-driven rather than Torvalds developing it based solely on his expertise and intuition.

I think the seeing the underutilized commands and flags (with real data not just a hunch) would have helped identify where users were not understanding why they should use it, and could have helped refine the interface and docs to make it gradually more usable.

I mean no solution is perfect, and some underused things are just only sometimes extremely useful, but data used smartly is not a waste of time.