Hmm. I guess imagine Tom & Jerry and how the music is synced with their movements, just on a MUCH smaller scale.
Say there's a scene when the music smoothly transitions from happy to sad, and something in the game reflects that - or maybe a song is playing through the intro and it only goes to the title screen when the song 'kicks off' - all that has to be perfectly in sync or it will look bad / cause upcoming events to be out of sync as well.
Tracking down the milliseconds and executing the right event(s) based on that seems to be the only thing I can do, and it works pretty well provided there are no lag spikes / slow load times or something...but they are inevitable it seems.