Same here. To make a base text color more evident, the author smushed elements and lost information. There is no point to a base text color, and it's OK if everything is highlighted, as long as it is consistent.
If "everyone is getting syntax highlighting wrong", it's more likely that you're the one who is getting it wrong.
I do agree that it is harder to see the return misspelling with the more detailed syntax highlighting, but when I'm coding I'm mainly focused on the actual tokens I'm writing, and you better believe I pick up on if the syntax highlighting is showing my token as wrongly being a keyword rather than an identifier, or the wrong type of identifier. Even if I don't actually know what the colors are, it will look wrong.
Rainbow vomit syntax highlighting logically seems like a bad idea, but like many things, even though it's hard to prove that it's better, I think a lot of people will agree that it really is better. Especially with tree-sitter and LSPs giving more detailed syntax highlighting; I can feel the benefits of 'oh, that's the wrong identifier color, isn't it?' to the point where if I run into an edge case that breaks highlighting it becomes immediately noticeable.