Aphyr has a distinguished track record of breaking things that have gone through "years of thinking and review" and "months of testing". I'll trust his analysis over anyone else.
Personally, I would welcome Aphyr trying to break the Kafka EOS guarantees. Like all software, there will be bugs, and the exposure will only make it stronger and more viable.
I write as one of the implementors of exactly once guarantees in Kafka.
I write as one of the implementors of exactly once guarantees in Kafka.