What would you check first if consumers are reading the same messages multiple times after a deployment?

Checklist: 1) Was the consumer group ID changed? A new group has no committed offsets, so it starts from auto.offset.reset (default latest; earliest would re-read the whole log). 2) Were offsets reset during the deployment, for example a script running kafka-consumer-groups.sh --reset-offsets? 3) Commit timing: with enable.auto.commit (default true) offsets are committed periodically in the background, and with manual commits a bug or a commit after a restart means records processed since the last commit are redelivered. 4) Rebalances during a rolling deployment: when partitions move, records processed but not yet committed are redelivered; exceeding max.poll.interval.ms (default 300000) also triggers one. 5) Was the topic recreated? Offsets committed for the old topic do not apply to a new one. 6) Did the group sit empty longer than offsets.retention.minutes (default 10080, 7 days)? Its committed offsets are then removed and it restarts from auto.offset.reset. 7) Describe the group with kafka-consumer-groups.sh --describe to compare CURRENT-OFFSET and LOG-END-OFFSET. In practice, a group ID change or an offset reset in a deployment script are the first things to rule out.

Expected key concepts: group ID changed, auto.offset.reset, reset-offsets, enable.auto.commit, committed offsets, rebalance, max.poll.interval.ms, offsets.retention.minutes, topic recreated, kafka-consumer-groups.sh, deployment script