The ExecSummary does not report the time spent on the network. The time reported on the EXCHANGE node is misleading and does not count network time. This makes it impossible to identify the performance bottleneck.
The ExecSummary reports the non-waiting time spent in the EXCHANGE as the network time. Obviously, the active time spent in the EXCHANGE is zero (or close to zero). It should be replaced by the corresponding Data Stream Sender total time instead.
Examine the corresponding time in the Data Stream Sender instead.