Comment by late_night_fix

Trying to enforce "deep understanding" is optimizing wrong layer. Focus on: . Fast rollback . Good observability . Small PRs Understanding become emergent, not enforced.