Recently I've reviewed a few MRs that ended in me writing more comprehensive guidelines for the project.

The author hints at the bidirectional aspect of code review, but they miss that each MR is teaching you how your contributors are getting confused.