Trace exhibiting odd flatspot in NDT 7 pcap

[mattmathis]

Pcap for ndt-nnwk2_1611335823_00000000000C2DF0, suggests that the thread writing data stalled for ~250 mS, right at the end of the test (between 10.1 and 10.35 seconds).

This may be after the point where the original test computed the average rate (right at 10.1 seconds) so it might not affect the original measurement, but it complicates any post processing, since the data has to be cut off at 10 seconds, even though the test ran longer.

[stephen-soltesz]

A general problem here is mapping events (e.g. start, stop, close) from a primary measurement (e.g. NDT download) to a sidecar datatype (e.g. pcap, tcpinfo, traces), which may have a different set of events (earlier or later start time, later close times).