Three things get confused and are worth separating. Correction fixes the immediate instance: reject the batch, replace the gasket. Corrective action removes the cause so it does not happen again. Preventive action addresses a cause that has not yet produced a failure, usually identified from trending or from a problem seen elsewhere. Most sites do correction well, corrective action inconsistently, and preventive action rarely.
The step that decides everything is the effectiveness check, and it has to be defined before implementation rather than reverse-engineered at closure. 'Effectiveness confirmed' with no pre-defined measurable criterion is not evidence, and it is straightforward for an inspector to test: ask what the criterion was, when it was set, and what data was compared against it.
Retraining as a corrective action deserves particular scrutiny. Where a process invites an error, training a person not to make it is a delay rather than a correction. A deviation log where retraining appears repeatedly against similar events is a documented pattern of ineffective CAPA — which is read as a statement about the quality system, not about the individual events.
Backlogs are an arithmetic problem, not a discipline problem. Events arrive faster than they close, and teams then close the easiest items to show progress, leaving the queue weighted towards the difficult, high-exposure items that should have gone first. Triage by regulatory exposure before touching anything.