Correct is where the controller decides what to do about the gap that sense measured. Its rule is narrow on purpose: each cycle addresses the single largest deviation from the setpoint, then stops and senses again. It does not try to fix everything in one pass.
This is the discipline of a controller holding a setpoint rather than a script running a checklist. A system that chased every deviation at once would overshoot and oscillate, the way an over-eager control system hunts around its target instead of settling on it. Taking the biggest error first, one correction per cycle, is what lets the loop converge instead of thrash.
The correction itself is a decision, not the work. Having chosen the largest deviation, the controller hands the actual labour to dispatch; it never writes application code with its own hands.