Hello, I had a question similar to https://apache-flink.slack.com/archives/C03JKTFFX0S/p1674668351943749. Are there any potential upcoming plans to support this? I also found FLIP-186 which had a proposal for the issue, but from what I can find there hasn't been a lot of detailed discussions or started implementation. My use case is doing a join on two high throughput kafka topics. Even with the data partitioned on the incoming topics, the

keyBy

is resulting in hundreds of mb/sec of network shuffling, so we could see a significant performance bump if we could better preserve the upstream partitioning in Flink