That assumes the selection process is effective, but FAANGs are all monopolies so there's very weak selection pressure for them to distinguish candidates correctly. They don't get punished for their mistakes.
I'd argue the people at FAANGs are the people who are best at hitting the requirements of a FAANG. That's still a competition and they'll brain drain from smaller companies, but it creates arbitrage on unrecognised ability and other opportunities may exploit that (e.g. starting your own venture relies much more on genuine ability than where you stand on a FAANG progression matrix).
I was contributing to an open-source FAANG project recently and the oversights they'd made in the codebase were surprising to me. There were very basic errors and mistakes and the people managing the codebase didn't understand what some of the functions were even supposed to do. The API wasn't consistent. Tests were testing behaviour that was outright wrong (and, somehow, passing). Feature prioritization was way off. There were half-hearted implementations of some features that were bad enough that they weren't useful. They weren't talking to other departments working on the same basic area (who are also open-source, and hold the state of the art), so there were important features that weren't ported. The project was small, too (but not unimportant).
I don't have high market value, but I was fixing some of these bugs in minutes. The guys managing that project are obviously a lot better than me overall, but I think it betrays a flaw in the institutional structure. Those weaknesses in the codebase would be cheap to fix, the problem is the FAANG standards are such that they are able to go unrecognised.