The KafkaExporter container consumes a significant amount of CPU on a cluster with thousands of topic/partitions.

Having an unbounded parallelism in fetching the information about topics [0] does not scale well. Every goroutine is a thread and if you have lots of topics, this will result in lots of threads, all competing for CPU time.

[0] https://github.com/danielqsj/kafka_exporter/blob/master/kafka_exporter.go#L369