Every serious academic knows it is complete crap, the problem is finding a better metric. Although the crappyness competes with "no metric at all" and I think the latter would be superior.
The only advantage the current system has is that it is stupendously simple, so at least it is hard to manipulate.
Institutional design needs legibility to track people's reputation at scale. The problem is that these metrics are often poor substitutes.
The difficulty is that nobody knows how to. My guess is that if you want good signaling, you'll need to find something that is difficult to fake. My guess are evolving benchmarks that measure many things in multiple dimensions, but benchmarks were easily gamed.
You can have it without benchmarks that can be gamed, but then it's basically down to "this feels right" and you have to trust the leadership to not be discriminatory, etc.
The only advantage the current system has is that it is stupendously simple, so at least it is hard to manipulate.