I don't think it has anything to do with v3. The audio overlay code is all in the kernel and all it does is fade the music down and mix it with the PCM sound (coming from flite in this case.) Because the flite sound is at full volume and the music's faded down, it's going to be comparatively louder than the music.

If flite has an option for changing the volume of the voice, that might be a good thing to try. I can't remember if the player's volume knob also brings down the volume of the audio overlay samples, if not, maybe someone has to work on a more robust way to do the audio overlay stuff. I did at one time attempt to play with the sample mixing code to scale down the PCM sound by a configurable amount, but I have no ARM assembly experience so my feeble attempts came up short.