Right. Most network interface *chips* are black boxes. It is hard to get any data that tells you what the timing relationship is between "twiddling a register" and the data appearing on the wire (or, conversely, between the data being on the wire and a register/interrupt reflecting its presence)
Not necessarily. The switch may not pass the traffic that A and B are exchanging to the machine "C". And, even if it did, there is no guarantee that the delay through the switch to C will track the delays that A and B experience (since the switch has an elastic store within)
Agreed. At the very least, I will be able to get a feel for the jitter that typical PC hardware/software experience as I will *know* what the real values seen by my boxes are and can correlate that with the *reported* observations on "machine C".
--don