Author here. The argument was that by numerological coincidence, a couple of very important numbers (world population, written characters, seconds in an epoch, and plausible process memory usage) just happen to lie right near 2^16 / 2^32. I couldn't think of equally important numbers (for a computer) near ~260k or ~64B. We just got unlucky with the choice of 8-bit bytes.
We need to be better at estimating require sizes, not trying to trick ourselves into accomplishing that by slipping in an extra bit to our bytes.