> NO one is talking about their intervention stats.

Interventions are a term of art, i.e. it has a specific technical meaning in self-driving. A human taking timely action to prevent a bad outcome the system was creating, not taking action to get unstuck.

> IF they were good I would assume that someone would publish them for marketing reasons.

I think there's an interesting lens to look at it in: remote interventions are massively disruptive, the car goes into a specific mode and support calls in to check in with the passenger.

It's baked into UX judgement, it's not really something a specific number would shed more light on.

If there was a significant problem with this, it would be well-known given the scale they operate at now.