Apache Flink

hi :wave: we are deploying Flink in `CLAIM` mode w/ <https://flink.apache.org/2018/01/30/managing-large-state-in-apache-flink-an-intro-to-incremental-checkpointing/#restoring-checkpoints-and-performance-considerations|incremental checkpoints> (in S3) `state.checkpoints.num-retained: 50` and notice that it takes several minutes for Flink to verify the retained checkpoints before job manager starts to schedule a job. The jobs of JM suggests its validating checkpoints sequentially. Does anyone have ideas on what we can try to speed up this step? We've tried allocating more CPU, IO and Network capacity to the job manager instance but that doesn't seem to help.