I've written a custom class that creates a fixed number of audio sources. When a new sound is played, it goes through the class, which creates a queue of sounds that will be played during that frame. The sounds that are closer to the camera are given preference.
If new sounds arrive in the next frame, I have a complex set of rules that determines how to replace the old ones. Ideally, "big" or "important" sounds should not be replaced by small ones. Sound replacement is necessary since the game can be fast-paced at times, and should try to play new sounds by replacing old ones. Otherwise, there can be "silent" moments when an old sound is about to stop playing and isn't replaced right away by a new sound. The drawback of replacing old sounds right away is that there is a harsh transition from the old sound clip to the new one.
But I wonder if I could just remove that management logic altogether, and create audio sources on the fly for new sounds. I could give "important" sounds more priority (closer to 0 in the corresponding property) as opposed to less important ones, and let Unity take care of culling out sound effects that exceed the channel limit. The only drawback is that it requires many heap allocations.
I wonder what strategy people use here?