Also, it's possible that a meritocracy is doomed to be subverted in any given enterprise.
Yes, it's great to strive to eliminate one's own biases and "confounding factors" to make one's organization stronger, but most "meritocracy" thinking is about finding the right magic numbers to describe people and that is its own bias/set of confounding factors.
People are often more than just "value", they are part of relationship webs, and full of complex emotions. You generally can't just assume a person is an interchangeable widget/commodity in your system and you insert paycheck and receive some magic amount of "value" in exchange.
A person's productivity absolutely depends on things besides skill, work habits, and leadership ability, however you define them and however you try to metric their value. A person's productivity might be powered by finding that random coworker's leftover cake baking experiments in the break room on those Wednesdays when they need it most. Another's might be powered by interesting gossip. Other people's are tanked by the extra buzz of the a/c on days in the Summer above 80F. You can't predict when someone might have a huge break-up that ruins their productivity for weeks and you can't predict when someone has just the right sort of vacation that their productivity is amazing for the next year... and so on and so forth.
The very term "culture fit" alone, and how often it is used in "meritocracy" contexts, is a giant neon sign that meritocracy doesn't exist in the real world and people are trying to keep themselves warm at night rationalizing that the decisions they make on semi-random metrics or gut "culture fit" decisions are the best they could make, rather than deal with messy subjective world of dealing with actual people.
I do share concerns with you about the ideas you are disadvocating. Such as that a person's productivity not a neat weighted sum. Such as that we could systematically find all the numbers that need to be summed. Plus concerns I'm not sure if you're getting at, such as the idea that sort of results from pushing a meritocratic view, which is that a person's value in the world is their value to some small collection of enterprises, their one job or their two jobs.
If you are paying people for their work and also providing some encouraging words, what do you then use to determine the pay packet or the words to say?
It's very easy to build a culture based on merit metrics and tautologize that numbers are objective, this merit system is based on numbers, therefore this merit system is objective and thus this is objectively a "meritocracy". There's a number of logical fallacies wound up in that thought process, but that's how a lot of bureaucracies get formed, and a how a lot of them rationalize themselves as objective/benign.
I guess the underlying problem is that merit is great, but the illusion that "merit" scales to form benign meritocracies is something that we need to question a lot more than we do.
We wouldn't build a bridge and assume our geometry is perfect. Bridges would fall down. We have to constantly try and measure and adjust for how closely our bridge adheres to the blueprint and adjust or refactor as we go.
Furthermore, how do you even define grades or metrics for "pure coding ability" or "works well in teams"?
Both have massive subjective components, don't they?
Isn't "works well in teams" really "works well on my team"? Trying to pretend like it's an objective measure of "works well in the platonic ideal of teams" doesn't actually make it an objective measurement.
Same with "pure" in "pure coding ability": adding an objective seeming word like "pure" doesn't excuse that there is no such thing as a platonic ideal of coding ability to measure against. There's "codes well in this problem domain", there's "codes well in this environment", there's "has the baseline experience to code well quickly now", there's "has the learning ability to get up to speed quickly", etc... Where's the objective measurements?
A lot of "meritocracy" talk is built on sandcastles of wish-it-were-objective-metrics without any sort of in depth analysis to how objective the metrics can possibly be, much less actually are in pragmatic, impure reality.