> There's an argument that serialism forces composers to use those operations instead of falling back towards conventional theory. But in fact the opposite happened. It didn't take long for serial composers to start mining serialism for tonal elements.

Nothing of this sort happened except in very exceptional cases where it's very obvious and clearly intentional, such as Alban Berg (who chose adversarially tonal tone rows). Besides you assume the only benefit you get from serialism is atonality but that makes no sense because many composers used tonal rows too (i.e. tone rows with 7 or 9 pitch classes in the Major or Minor system). Not only that but different composers used serialism for different purposes.

You are however right that Schoenberg assumed 12 tone serialism implies some form of atonality, which is incorrect. Research shows that he had a bias towards atonal tone rows.

Lerdahl's criticism of serialism is also convincing, i.e. the abstract structure of music does not translate to the heard structure of music. It's fair to classify this as the "art" part of writing serialistic pieces that theory is insufficient.

Besides all of this, traditional music theory teaching focuses too much on serialism as a "technique" and not enough as a "structure", which was the main point of Schoenberg. For example, Elliot Carter claims he doesn't intentionally use row vectors or row matrices in his compositions, but his works can certainly be analyzed and understood this way. It seems like for a minor subset of free atonal compositions serialism is the natural structure of this kind of music. This point is often missed.

I firmly disagree with your points about Schoenberg. I think he's one of the two gigantic innovators of music in 20th century: Philip Glass and him. (I understand this is extremely controversial, opinions are mine). It really takes a lot to make artistic statements so profound your works look misplaced in your canon but are still firmly part of it.