Yes, I managed to replicate the bug, but not to test the proposed workarounds.
I hoped that the "distinct snapshot length" setting could be set to unlimited, but now I realize that this is not possible.
As a consequence, the snapshot and the EOS are always sent to the session at once, hence the race condition that triggers the bug is always possible.

In practice, the only condition for the bug is the combination of DISTINCT item, unfiltered dispatching, and snapshot handling.
This is worse than I initially thought.
We will publish a new release with the fix as soon as possible and will let you know.