I think it really depends, background audio is not a big deal when it comes to networked gameplay, if player A is playing background audio 1, and player B is playing background audio 2, they won't actually care, this is similar to how 2 games can be completely out of sync animation wise across 2 games, or things like rag dolls will be out of sync, because they are not things that a player will notice.

Now lets look at something like gun shot sounds, explosions, characters giving orders, etc. These are things very important to gameplay especially in something meant to be tactical like the Battlefield games, if you don't here a bullet whip by your head, it gives the sniper on the other end the advantage of remaining unheard and unnoticed, events allow these sounds to be passed over the network from Server-Client and have each player who is within the audible location to here the sound.

This doesn't mean that background audio can't be dispatched by events though, if you want to dispatch an event at client start to synchronize background audio there is nothing wrong with this. Generally in most engines I have seen (as well as my own), there are different component types.

- Audio Listener component, this is generally attached along with a camera that the player views the game from, or maybe even forced along with a camera. This typically represents something that is able to receive audio events.
- Audio Source component, this is something that produces sound, and can represent either a globally playing soundtrack, a trigger-able sound, etc. Usually you can also fire off sounds without these audio source components, as they act as a simple way to attach an audio clip to a logical entity in game.

The way that you separate the audio between logic-view is up to you, but here are some points to think about.

- Audio listeners don't necessarily need to play audio, but could simply be a pass through to say that 'this game object cares about audio, so lets make sure any attached view knows about the sound'
- Audio listeners can be processed on the logic side, or the view side. If your AI are completely logic based this may be the only way to go. Otherwise you may need to dispatch Audio to the view.