Github user bowenli86 commented on the issue:
https://github.com/apache/flink/pull/6130
Not really. It's not about having n copies of data. One use case is
File-fed stream pipeline usually runs very fast with inadequate metrics, users
need to run it end-to-end for a longer time to gather stable metrics and tune
all components in the pipeline.
I'll close it if community is not interested.---
