We need to document how is pipelining across chain of relays impact end to end latency compared to full object assembly per hop.