Inferential queueing and speculative push for reducing critical communication latencies