I rebuilt an old dissertation idea as a C#/.NET benchmark to test whether Hilbert curve ordering helps solve Shikaku puzzles. Pure Hilbert ordering beats row-major scanning, but MRV is far stronger as a primary heuristic. Blind Hilbert tie-breaking helps often but has nasty outliers. The best result came from using spatial locality cautiously. Apply MRV first, then use local proximity near the most recently placed rectangle only when MRV leaves a tie.