Why can a handful of failures flood a dead letter queue with thousands of messages?

Failures rarely arrive one at a time — they arrive in batches. A single upstream problem (a bad schema deploy, a rotated credential, a config typo) can cause every record in every batch it touches to fail the same way, so one root cause is amplified into a huge pile of DLQ records. The fix isn't to read all of them: sample a handful, group by exception class, and you'll usually find one or two dominant failure modes hiding behind the volume.

Expected key concepts: dead letter queue, DLQ, batch amplification, exception class, poison pill, triage, root cause, Kafka Connect