So, between the new iOS rev and Construct 3 updates, the behavior of text-to-speech appears to have changed. It appears that I have to initialize the text engine by getting it to speak at least once in a user-initiated tap. Any idea when or where this new behavior is documented?