Resolved an issue where words at audio chunk boundaries could occasionally be lost during streaming transcription.
The correction targets the joins between successive audio chunks, helping words survive the transition during a longer streamed recording.