What would that process look like?
Perf reviews these days are basically, "Did you do some important/useful things? Did you tell other people at the company about those things so they'd benefit, too?"
And interviewing (other than the inane hoop-jumping of getting through the recruiters, and the subset of interview loops that focus on LeetCode) is largely, "Can you write a bit of decent-looking code? Can you design a globally distributed self-healing system without being a total buffoon? Can you demonstrate some semblance of introspection?"
There's some bureaucratic fluff surrounding it all at the larger FAANGs, but it does seem largely reasonable overall — especially if you (or your manager) are willing to cut through the red tape.
I could see a fair number of improvements and tweaks to be made, but... fundamentally different? You want to throw everything out and do it all differently?
Having different review processes adapted to different job roles that exist at different levels is the easiest thing to defend. Heck, given disparate impact considerations and the fact most systems will have incidental unequal impacts, failing to do so could have you fall out of the business necessity criteria into unlawful discrimination at the levels the process isn't adapted to.
Are the “obvious reasons” that, despite the practical and legal benefits, someone would have to expend effort designing the appropriate system?
L9 is conservatively making $2 million / year so the price of getting perf wrong is high.