The three oversight metrics are the part I would steal for smaller teams: coverage, review latency, and escalation rate. Most small shops only track whether the work shipped, so a model can look productive while nobody checks which artifacts it actually touched. The one practice I would add is classifying work by reversibility before it starts. A reversible draft gets a light check; anything that writes, sends, or spends gets the full replication and a second pair of eyes. And the counterweight point matters most: the moment a system starts grading its own work, the oversight has to come from outside that system, or it is just a progress report wearing a reviewer's badge.