Hi all, we're seeing an inconsistent behavior in Flink committing Kafka offsets. Flink job lag shows a different number than what is committed to Kafka. Below are two screenshots that show lag seen by Flink vs Kafka. Has anyone faced this issue before? Our assumption: This inconsistency might be due to the fact that partition leaders might have moved to different brokers on MSK and Flink now has stale information about them.