Total consumer lag is climbing but only one partition is behind, and adding consumers changed nothing. Why?

One partition has one consumer, by design. If the key distribution sends a third of the traffic to one partition, the parallelism you bought is unavailable to the work that needs it.

Grounded in researched DevOps, SRE and platform engineering interview loops, written to a senior-engineer editorial bar, and never padded to hit a word count.