I'm pretty sure that problem is solved from a technical point of view - though it might not be commercially available yet due to economic and regulatory reasons.

It has very much not been. None of the players on the market are able to do anything close to this. You either have self-driving on carefully pre-mapped routes, with on-call human suport ready to jump in for more complex cases (Waymo), self-driving in very specific conditions with few seconds notice to the human driver (highways, with speeds below 60 - Mercedes), or not self driving, where the human driver has to be ready to intervene with no notice whatsoever (Tesla).

You're trying to move the goalpost from "self-driving" to "self-driving in scenarios where the route has been fully premapped".

Might as well move it to "as long as there's no semi crossing the road" to cover for Tesla.

Centrally-managed clean and rigid setting is not the full extent of the problem domain.

On the contrary, I was saying that self-driving in scenarios where the road has not been fully premapped is solved.

Meanwhile, all the self-driving that's currently happening is within well-controlled small geographic areas and some limited interstates/highways, and based on matching observed surroundings to pre-made very detailed maps. Okay, sure.

As I said, I'll believe it's solved when these limitations are no longer necessary for it to work. You can choose to believe it based on a research project or marketing materials.