Howdy! It seems like `requestDequeueTimeNanos` is ignored when handling an envelope request, which could lead to `requestDequeueTimeNanos` ending with "-1", so that the metric,`LocalTimeMs` is not correct.
I created an issue related to this together with a tiny PR. issue: https://issues.apache.org/jira/browse/KAFKA-14032 PR: https://github.com/apache/kafka/pull/12360 <https://github.com/apache/kafka/pull/12360> I'd appreciate it if someone is interested in it. Feiyan Yu