Consumers Under Fire — Week 2 CheckWeek 2 of Kafka in Production: the consumer-group failure chain, the rebalance protocol's evolution, and cross-cluster failover. Every question traces to a case study you read this week.

A consumer group stops processing with zero errors in the application logs. Per Adobe's production experience, what's the FIRST hypothesis to check?

Kafka runs two consumer liveness clocks. Which failure does max.poll.interval.ms catch that session.timeout.ms cannot?

In PayPal's load test, the consumer outran its BigQuery sink. Order the resulting cascade correctly:

What does incremental cooperative rebalancing (KIP-429) change versus the eager protocol?

Grab's Coban platform cut consumer resource waste ~45% and eliminated rebalance churn with one structural decision. Which?

Why are committed offsets meaningless on a replica cluster in multi-region Kafka?

After a regional failover, Uber's consumers resume from the SMALLEST offset mapped across regions. What does this policy deliberately trade?

What architectural change does KIP-848 make to the rebalance protocol?