Why do CS researchers not care much about progress on new problems?

There is an element of competition to it, and if there aren’t other people competing for the same thing then there isn’t any prize. For instance there are things like Kaggle, TREC and other competitions around putting up the best numbers for a particular problem which has the pernicious effect of creating “machine learning” experts that can do the MNIST digits to death but leave you wondering if the same methods would be successful for some similar problem (say handwritten Hiragana characters)

I think how new sports are frequently pioneered by somewhat older people who have aged out of being competitive at well-established sports and once there is a scene younger people take over. Younger people see jumping into an existing competitive hierarchy as the way forward, some older people see they can’t win at that and then start something new.

That's a distinction between basic and applied research. People who care about applied research either found a startup or join Google or Meta. That way they can get plenty of reward.