Github user fmthoma commented on the issue:
https://github.com/apache/flink/pull/6021
@tzulitai I added some docs.
As for the `flush()` vs. just waiting: As I see it, the
[`RecordMaxBufferedTime`](https://github.com/awslabs/amazon-kinesis-producer/blob/ce77505306c104a6016b0c081df4715d05ac9201/java/amazon-kinesis-producer-sample/default_config.properties#L239)
option (default: 100 milliseconds) limits the time a record should be kept in
the queue in the absence of pressure. Hence I think that the `flush()` is
indeed not necessary, unless a user purposefully sets the `queueLimit` too low
*and* the `RecordMaxBufferedTime` too high.
Also, I added [another
comment](https://github.com/apache/flink/pull/6021#discussion_r190154347)
concerning the `sleep` vs `wait`, that github unfortunately displays as
»outdated«.
---
