Exception with AVFoundation when using Speech framework in background (sporadic behavior)

My app is a VOIP/GPS app and I'm trying to use the Speech framework in the background. The behavior is actually quite sporadic. Sometimes it works and sometimes it doesn't, in that I get this exception:

*** Terminating app due to uncaught exception 'com.apple.coreaudio.avfaudio', reason: 'required condition is false: IsFormatSampleRateAndChannelCountValid(format)'

Let's assume that I have user permission and that the speech recognizer is available (both are true at the time of the crash). The method I invoke is:

private func listen() throws { 
    if let recognitionTask = self.recognitionTask { 
      recognitionTask.cancel() 
      self.recognitionTask = nil 
    } 
    self.recognitionRequest = SFSpeechAudioBufferRecognitionRequest() 
    guard let recognitionRequest = self.recognitionRequest else { fatalError("Unable to created a SFSpeechAudioBufferRecognitionRequest object") } 
    guard let inputNode = audioEngine.inputNode else { fatalError("Audio engine has no input node") } 
    recognitionRequest.shouldReportPartialResults = true 

    self.recognitionTask = speechRecognizer?.recognitionTaskWithRequest(recognitionRequest, resultHandler: { (result, error) in 
        // Code here in which I look for certain phrases - irrelevant 
    })  // EXCEPTION BREAKPOINT LANDS HERE

    let recordingFormat = inputNode.outputFormatForBus(0) 
    inputNode.installTapOnBus(0, bufferSize: 2048, format: recordingFormat, block: { (buffer: AVAudioPCMBuffer, when: AVAudioTime) in 
      recognitionRequest.appendAudioPCMBuffer(buffer) 
    }) 

    self.audioEngine.prepare() 
    try self.audioEngine.start() 

}

Most of these variables are declared or defined outside of the method body:

  private let audioEngine = AVAudioEngine() 
  private var speechRecognizer: SFSpeechRecognizer? 
  private var recognitionRequest: SFSpeechAudioBufferRecognitionRequest? 
  private var recognitionTask: SFSpeechRecognitionTask?

Any ideas on why this crash occurs? The issue clearly isn't to do with the Speech framework but rather with AVFoundation when creating a live recording session.