Hello, I am using Table API with session windows for one of my pipelines(with some aggregation logic). I am seeing a weird issue, that suddenly there is a huge spike in the job output rows per second, and then it falls to 0. At the same time I am seeing the checkpointing size also drop to 0. And it fails to recover beyond that point unless the app is restarted. This leads to data loss in the pipeline. I am not seeing any errors in the logs though. Any pointers regarding a potential rootcause would greatly help, thanks.