AVSpeechSynthesizer, usesApplicationAudioSession, ducking delay

In iOS 13 there is a new property on AVSpeechSynthesizer: usesApplicationAudioSession. If I set it to false, then I get a nice behavior that I've wanted for a while -- the AVSpeechSynthesizer lowers the volume of music playing in the background, while it speaks. The only thing I don't like is that the music volume remains lowered for about two seconds after my synth speaks a word. Are others seeing this? Is there a way to decrease this delay to almost zero? I checked my AVSpeechUtterance, and it's `postUtteranceDelay` is zero.