ARCHITECTURE Signal 375
Retries don't fix eventual consistency
Engineers building distributed systems often default to retries and dead-letter queues when dependent data hasn't arrived yet, creating unnecessary operational burden and ordering problems. Recognizing that missing data under eventual consistency is a valid system state—not an error—leads to simpler architectures where events wait for their prerequisites naturally.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Missing data in an eventually consistent system is a normal state, not a failure condition, so retries and dead-letter queues are the wrong response.
Storing each incoming event and checking whether all required pieces are present before processing removes the need for retry logic and manual message replay.
Retries remain appropriate for transient infrastructure failures like network timeouts or dropped packets, but multiplying retries without evidence of recovery suggests the wrong problem is being solved.
THE CLUSTER