It's been requested that we enable a batching mechanism to let users of the API tell us to process a batch of frames at a time (rather than the lowest latency we can support on a given system), trading off higher latency for applications that would rather minimize power (e.g. a simple background-music-playback app).