Sorting Networks and SIMD — How Columnar Databases Sort in Memory

This section is for when an interviewer asks 'how would you make this sort 10x faster at the hardware level?' or 'how does DuckDB sort so fast?' The answer involves sorting networks, SIMD instructions, and cache-aware algorithm design. This is principal engineer territory — you won't implement this in an interview, but knowing it exists and why it matters demonstrates systems depth.