If BAL contains the read locations, we can implement a solution that reads all BAL locations at the start of block processing and applies them to the accumulator. This approach would allow us to maximize read parallelization and improve transaction execution performance.