为什么与分片集群中相应的mongos相比，MongoDB中的各个分片报告的删除操作更多？

So I have a production sharded MongoDB cluster that has 8 shards (replica sets) managed by mongos. Let's say I have 20 servers which are running my application and each of the servers runs a mongos process that manages the 8 shards.So I have a production sharded MongoDB cluster