[majormajor]

I actually think all three of the things you list as "unintentionally testing for" are extremely important.

Systems often work well for months and don't need touching, then have an incident, or need a new change, etc. You can remember that stuff fairly quickly? Perfect. You can't? I think that's a potential flag, directly related to "able to identify issues in a complex system" and "able to figure out solutions to said problem."

But the other reason why not many candidates do well on that is because there's also just a lot of companies where the day to day work is not-complex problems on not-complex systems (even (especially?) at FAANG where the typical dev is a very tiny piece of the machine). So in that case we have to fall back to made-up approximations for "complex" problems that don't require a ton of extra context.