So here's my question: how useful is h-99 (are they overrated as learning
tools)? I find myself solve most of them in a "from the scratch" fashion
(e.g., no Monad, no Applicative, no Functor aside from List and a few
Maybe). Some of them are paper-worthy, for example the prime problems. I
hope some guru-level Haskeller could do away the missing few, and maybe
dive deeper into the surface to produce more insights (like the knights
travel page or the sieve paper, which are both beautiful).

As another new-haskeller I say: Yes & No.

"Yes" for necessity of several described algorithms and gained intuition for real programming.

"No" for needed ammount of work to understand pitfalls of Haskell when real programming.

And of-course, problems and solutions are not annotated with there typical real world aplications, they are not obvious for average beginners. Why would I make Binary tree balanced, when I don't know, what I'll gain (except balanced binary tree)?