Diagnose and recover from ProducerFencedException in a Kafka exactly-once pipeline

Steps

1. Recognize that ProducerFencedException means Kafka detected a newer producer instance registered under the same transactional.id — every initTransactions() call bumps an internal epoch, and any producer still using an older epoch is fenced as a zombie.
2. Treat it as fatal and non-recoverable for that producer instance — don't retry on the same KafkaProducer object.
3. Recover by closing the fenced producer, constructing a brand-new KafkaProducer with the same transactional.id, and calling initTransactions() once before any beginTransaction().
4. Distinguish it from InvalidTxnTimeoutException, which fires when a producer's transaction.timeout.ms (client-side, default 60000 ms) exceeds the broker's transaction.max.timeout.ms (default 900000 ms) — that's a config mismatch, not a zombie-producer event.
5. Audit for accidental duplicate producer instances sharing a transactional.id (e.g. two replicas briefly running during a rolling deploy) — that's the most common real-world cause of fencing.