Handle SCN gap

We deployed debezium in our testing env, normally the SCN increase several thousands in one query loop, debezium works well. But sometimes the SCN increased much bigger in a one query loop, in the following example, the SCN increased by 2219052488. In fact the row count during that time are similar with the other normal query loop, it's just a SCN gap.

Although we set the  "log.mining.batch.size.max":"200000", it took a long time to catch up when the SCN gap happen. 
 prevEndScn 17800237736902, time=2021-09-23T05:27:03Z current SCN 17802456789390, time=2021-09-23T05:27:09Z

We can detect such a case, and use the current SCN as end SCN, thoughts?