'It worked after clearing cache' is not a root cause

Situation

Complex CMS platforms have several cache layers, and `drush cr` genuinely fixes many visible problems. The danger is turning it into a reflex for every strange behaviour. State-management defects then remain invisible and return later, long after the original context has been forgotten.

Approach

When a cache rebuild removed the symptom, I looked back at the deployment, config import, container rebuild or data change that preceded it. Missing invalidation was treated differently from genuinely corrupted cache state. The goal was to make required invalidation part of the next release path rather than a manual production afterthought.

Outcome

Cache clearing remained an important recovery tool without becoming the diagnosis itself. Several deployment sequences became more explicit, and fewer production runbooks ended with the vague instruction to clear cache whenever something looked strange.