Unfortunately, you can use this to send thousands of bits between syscalls, so the simplest error correction would fix that, with very little effort or overhead.
The demo already uses error correction (I'm not sure exactly what causes the errors, but I'm guessing the processes sometimes end up briefly scheduled on the other core cluster)