Hi , we have issue https://github.com/apache/pulsar/issues/22601 , I have tried to understand the issue but it's difficult, does anyone could help us debugging online by zoom? I could show how it happen in our preproduction environment. I have tried to reproduce in pulsar-perf produce • run in jdk21 • run in high concurrent perf producer. • also try to upgrade bookkeeper to 4.17.0 but it could only reproduce in our producer which consume data from kafka and publish that message to pulsar about 6000 qps I can't reproduce by pulsar-perf tool in the same qps and random big payload. and the issue seem disappear when we remove the payload > 1.5K bytes