Each style has its place. Just like you, I found the "harder" problem sets and related classes much more interesting and exciting. That significantly correlates with me being in this field after all. That is also a poor style to pick for an intro class, as the intro class needs to provide an overview of the field and teach many diverse skills in relatively minimal depth. The intro classes are not meant to teach mastery. Good schools pay sufficient attention to their students to bump up the occasional student that is taking the intro class instead of the more appropriate for them mastery class.
Having a grading philosophy that boils down to pass/fail A-F distributions is great from the perspective that there are students who paid for a class and the professor ensured that they had acceptable knowledge by the end of the course. The wide grade distributions tended to be hated by most students, but the few folks who scored well could make up for bad homeworks etc.
Now the curious thing about these two grading approaches is that I always recall the material of the classes with wide grade distributions with more fidelity than those graded on 100% or bust curves (which is saying something after 11 years). In hindsight I believe the "harder" problems sets lead to deeper learning as once could always see what the next skill progression was and incentivized taking risks on spending an extra few hours studying or working a problem to try and get a better score. I also wasn't afraid of coming up with a wrong answer through brave effort vs. looking for a pre-baked solution.
The Featured Article's core claim is that by adopting the 100% or bust style of grading that students optimize to avoid errors. This strongly incentivizes students to crib off of each others work, past homeworks, and other online sources to ensure that they've identified a truly correct answer. It also forces professors to keep problems closer to the book/lectures such that students are largely guaranteed success given some amount of student effort.
It could be that the older method of grading wasn't doing its job by selecting geniuses - but by incentivizing students to take risks. Is it that surprising that students who spent 16 years of school optimizing to avoid mistakes are extremely risk adverse?