What are the challenges involved in making a delay seem intentional and not a mistake or a bug?

Those challenges are absolutely significant. One of the first things that came up when we first implemented a prototype of this was that when you walk away from somebody speaking and you start to see the lip sync go out, that's kind of like, it's realistic, but people have seen movies oft enough that they don't expect it. So we have to strike a balance between realism and the experience people are a little more used to on screen. There's also the challenge of camera zoom, that's huge. So when the camera's zoomed in and the audio listener is actually significantly further away than it appears on screen, that causes problems, so we're having to overcome some of that as well. But yeah, plenty of challenges.