Thing is, current flashcard software (including Anki) does not support complex "study and understand" workflows very well, but a lot could be done to improve it. Even something as trivial as supporting explicit dependencies among cards, i.e. "don't bother to show card X unless you have managed to push out repetitions of cards Y and Z beyond $time_interval" would help a lot.

This would also enable flashcard/SRS workflows to interoperate seamlessly with other memorization techniques, like memory palaces or mind maps. Foreign language vocabulary is the rare example of a domain that doesn't need these things, so flashcards are perceived as "good enough" for it.