Hiiiii ! I am trying to troubleshoot what seems to be slow event time temporal join using FlinkSQL. I am recording raw trades and quotes from nasdaq in 2 different topic. I am augmenting the trades with the quotes using event time join and I am trying to understand why I am observing such a high latency (~400ms in median/avg) between the Kafka timestamp set by the sink task and the kafka timestamp set by the raw trade producer. I have excluded a watermark issue by setting the watermark on the right side of the join to infinity. Does somebody know what is the expected latency of the event time join in Flink SQL ?