I haven't done the stereo tool version of it, but I once used a TBC to delay 25 frames to match a line in- line out delay of 0,5 secs on a windows XP Graphedit graph loading a directx plugin as an audio maximiser.
That was back in the analog TV times, so today that we went all digital you could just preset your switcher to buffer 1 second of video (if that setting is available) and then use the whole delay as displayed into sound cards section of ST to match it to 1 second and a little longer (remember that light speed is enormously higher than sound speed, so our brain expects audio to reach some fractions of a moment after lips move, that information is essential to all lip sync scenarios. There are even some MPEG encoders that can handle this for you, but most of them have settings of delaying audio due to the reasons stated above. in ST version, our audio is already delayed so we need to delay video to match, which is costly in terms of RAM when you need to buffer HD video of half a second in SDI uncompressed while the audio path does its magic.
a delay of 5 to 10 mSec is expected from viewer side. Most TV sets won't implement that delay, as TV set manufacturers expect TV stations to handle the scenario.
TRY this at work, please post results. Do it with some video editing software, preferably on TV News / anchor content.
You will be amazed of how unnatural will look, having audio before video, and how far later you will be able to "draw" the audio before it seems unnatural.