The review process did everything right.

Someone opened the pull request. Someone else read through it carefully. The logic was sound. The tests passed. The naming was reasonable. Nothing obviously wrong. Approved.

Three months later that feature is the one nobody wants to touch. The patterns it introduced do not match anything around it. The state management approach it used became a local convention that contradicts the one two features over. The component boundaries made sense in isolation and make no sense in context.

The review did not fail. It caught what it was designed to catch. The problem is that what it was designed to catch does not include missing AI standards.

What code review is actually for