Comment by rvnx

This shows the failure of human reviews alone, an LLM-based reviewer would have caught it. Both approaches are complementary