This should be simple to implement. If, for any reason, re-muxing fails, we should be able to retry with re-encoding instead.