Was hoping for 50%, but if 64% holds up that's still good, a reduction from 6 to 4 hours for sequential read. The remaining question is how much performance on a memory constrained machine will be impacted by reading txs that have been written out of order.