I see. It seems that the two first events, the snapshot and the EOS, are enough to trigger the race condition (at 03:13:19,821) and the bug.
This means that, unfortunately, the conditions for the bug are broader.
Please try removing the "distinct snapshot length" setting; it is configured in adapters.xml. It still causes the two events to be dispatched to the session together.
If you still have problems, please also try enforcing a short delay before sending the EOS.