consuming hangs at empty offset caused by aborted transaction

[rickif]

When there are aborted Kafka transactions in the topic, the messages may be with discontinuous offsets.
Suppose we have kafka message as following:

offset 10: with message
offset 11: with message
offset 12: empty message caused by aborted transaction 
offset 13: empty message caused by aborted transaction 

The query_watermark_offsets would return low/high as 10/13.
But, the Message *consume (int timeout_ms) would hangs at offset 12 and returns RdKafka::ERR__TIMED_OUT.
What I expect is that the consume could reach the offset 13 and return empty message.
Is it by design or something wrong with my config?

[edenhill]

The abort marker messages (12 and 13) will be ignored and not passed to the application, so unless there are more messages being produced to the partition the consumer will not return any new messages (and thus time out according to your timeout argument)