Consumer lag compares the partition log end with a group's committed next offset. It says how much log progress remains, not how long a particular business action will take. A group with a bad commit policy can report zero lag while business work remains incomplete. A broker failure and a slow sink also need different responses. These exercises ask for a metrics snapshot before any repair so the original backlog remains visible. Inspect the group identity and offset arithmetic. Moving offsets to the end makes a lag chart look better without processing records. A useful incident report preserves the original evidence and states what was completed.
Inspect the topics, partition logs and group offsets before changing the cluster.