Overloading and templates. I can now use unsigned char, uint8_t and byte as distinct types, meaning they can be separately overloaded and used as separate template specialisations.

That's not a purely hypothetical point; I already create custom types to do this. Not every 8-bit type is a character, nor is it necessarily an integer. I always found it frustrating that the default stream output was a character when using numerical quantities; now we can specialise raw output accordingly.