I'd say unless you've got a good manager, really good hiring, and are keeping close tabs on the technical people (just to make clear, I'm one of them those technical people) I figure you've at least got 1 our of 20 you should just be sacking on principle.
Focusing on productivity is of course the right metric, but the metric that gets used in practice is “Which of you helped advance your manager’s career?” After all, your manager is the one who decides who goes, and there’s no incentive to keep someone that does nothing for them over someone who helps them look good to the VP — which is who decides whether your manager goes. Same rule applies recursively.
You're left with middling who might have a hard time finding other jobs but could play the ladder game well enough - potentially with devastating consequence for health of the technical stack as more "features" get developed.
The skills one develops on the job and the skills one needs to actually perform well at a job are not really tested well by the standard big tech interview loop (startups can be innovative here, but it can go either way).
Generally the competent folks reluctantly acquire the silly skills required to pass through the gatekeepers though.