The test server could add some extra timing information to its response headers:

• absolute time of request start from the test server PoV. (Probably only useful in tests where the client and server run on the same machine.
• latency between the test server's obervations of request start and its own reply time. This gets a sense of time spend generating a reply.
• latencies between consecutive inbound request start times.

Nighthawk's client could then add opt-in tracking for these values in new histograms with relatively little effort.

Having insight into these timings may be valuable as they assist in tracking (distortion of) request-release timings as well as give some more insight into components that make up the overall latency profiles.