They should keep the test, but use a threshold (per major and school) with a lottery for everyone above the threshold. And admit a small number below at random to keep evaluating what the threshold should be.
Not only do lotteries mitigate Goodhart's Law (which rewards 'bad' candidates who appear good, whilst punishing 'good' candidates who don't appear that way); but even a perfect, un-gameable measure can waste resources chasing diminishing returns. Even something as trivial as spelling mistakes can be decisive; and whilst those judging might only care a little about spelling, candidates care a lot about winning, so it's in their interest to obsess over even such minor things.
This imposes a massive opportunity cost, whether it's spending whole childhoods in 'cram schools'; or university salaries going to 'grant proposal writing services' rather than educators and researchers; etc.
Except for the gigantic study commissioned by the UC system the article references?
“ Second, while high-school GPA has been found to be more predictive of success at college than standardized test scores at some schools, the exact opposite turns out to be true for students at UC schools. There, standardized test scores say more about which applicants are likely to earn a degree and to do it in less than eight years; they also correlate strongly with students’ GPA at the university.”