As a part of my PhD research, I have been working on few optimization algorithms which try to jointly optimize delay and traffic (WAN traffic) in a geo-distributed streaming analytics setting. So, to show that the optimization actually works in real life, I am trying to implement these optimization algorithms on top of Apache Flink. For emulating a real life example, I need to generate a stream of records with some realistic delay (order of microseconds for fast incoming stream) between any two records. This stream will then by ingested and processed by Flink.
Using the timestamps as is, in the form of event timestamps, only proves the algorithms from a theoretical/simulation perspective.
Hope this answers your question to some extent at least. Let me know.