iOS text-to-speech in background - ios

Question

I am having an intermittent (aargh!) problem playing Text-to-Speech in the background, triggered from Apple Watch. I have properly set up the background mode, the AVSession category, and the WatchKitExtensionRequest handler. (See below.) I had this working before, and can't figure out what changed. (Could it be iOS 9 has issues? "Before" means, among other things, iOS 8.)
The problem is this: when the app gets the request from the Watch and the app is either in the background or the phone is sleeping (locked), the speech sometimes plays right away, and other times doesn't play until the app is brought to the foreground. The OS seems to be sometimes queuing the audio, and sometimes not. I can't find any common thread between success and failure cases. I can verify with logging that the call to speakUtterance() is being made in all situations. But its behavior varies, apparently randomly. The only clue is that it might be the case that the longer the app is in the background, the less likely it is to speak right away.
This is making me pull my hair out. Suggestions welcome.
In info.plist:
Required background modes: App plays audio or streams audio/video using AirPlay
In AppDelegate.application:didFinishLaunching:withOptions():
do {
try AVAudioSession.sharedInstance().setCategory(
AVAudioSessionCategoryPlayback,
withOptions:.DuckOthers
)
try AVAudioSession.sharedInstance().setActive(true)
} catch let error as NSError {
// etc...
}
In AppDelegate.application:handleWatchKitExtensionRequest...():
var bgTaskId:UIBackgroundTaskIdentifier = 0
bgTaskId = application.beginBackgroundTaskWithName(
"Prose WKE handler",
expirationHandler: {
application.endBackgroundTask(bgTaskId)
}
)
//... Post notification to call Text-to-Speech
application.endBackgroundTask(bgTaskId)

Answer 1

Here's a workaround: play a second snippet of sound (I used a half-second of silence), using AVAudioPlayer, right after the call to speakUtterance(), This seems to "jog the pipeline".