I knew that JSON wasn't ideal for large data blocks, but I'd never have guessed that something as simple as a custom data format and a split() function would be so much faster.