Apr 8, 2025

Also spent all of the slack time running syncs and table scans with various bloom and sieve configurations. In doing so I squeezed out some more performance by preserving entropy in bucket selection and filtering, reducing body searches (now down to 0.1% average over the point table).