Enno Runne

Our initial tests do not show an improvement in throughput for non-backpressureing commits.

I agree that these API docs should be published to akka.io. Before doing this, get @theotown in the loop to spare the remaining parts of his brain.

Thanks for reporting. I think the `consumer.position` call in `CommitRefreshing.Impl` needs to be guarded with try-catch. Are you aware that Kafka 2.1 fixed [the issue that required the commit refreshing](https://issues.apache.org/jira/browse/KAFKA-4682)?...

We'd like to understand if people still experience this with 0.20 and 0.21.

Thank you for this update. Anyone else?

Thank you for reporting that here. So some state in the actor or Kafka client seems to be lost upon wake-up in certain cases. **I'm not aware of a workaround...

It is considered production ready. We do not expect any significant changes and might call it 1.0 before the end of March 2019.

Thank you @ValueYouth, we haven't found any evidence that this is caused by Alpakka Kafka, so your hint might help others to identify the problem.

Thanks for the report. Interesting that the send callback is called as if it was successful.

Thank you for digging into this and sharing your findings. I agree it would be great to discuss the need for idempotence and acks. This does not only apply to...