Explain the Kafka consumer's heartbeat threading model and how it relates to max.poll.interval.ms.

With the classic group protocol, a consumer sends heartbeats to the Group Coordinator from a background thread at heartbeat.interval.ms (default 3000 ms), while the application thread calls poll(), which fetches records and handles rebalancing (KIP-62). The coordinator removes a member whose heartbeats stop for session.timeout.ms (default 45000 ms). max.poll.interval.ms (default 300000 ms) is a separate, locally enforced limit on the delay between calls to poll(): if it is exceeded, the consumer stops sending heartbeats and sends an explicit LeaveGroup request, so the group rebalances and reassigns its partitions even though heartbeats were healthy. This guards against a livelock, where a consumer keeps heartbeating but makes no progress on its records. The consumer rejoins the group on its next poll(). With the newer group.protocol=consumer, heartbeat.interval.ms and session.timeout.ms are not used; the broker-side group.consumer.* settings control them.

Expected key concepts: background thread, Group Coordinator, heartbeat.interval.ms, poll(), max.poll.interval.ms, session.timeout.ms, LeaveGroup, livelock, locally enforced, rejoin