为什么会出现单个node的indexing指标会比集群所有索引indexing指标的和还大的情况

在用[_nodes/stats?human=true]API看的时候，得到某个node的返回结果如下：

"nodeID":

    "indexing": {

        "index_total": 19137396947,

        "index_time": "18.7d",

        "index_time_in_millis": 1622193178,

        "index_current": 0,

        "index_failed": 58,

        "delete_total": 38,

        "delete_time": "41ms",

        "delete_time_in_millis": 41,

        "delete_current": 0,

        "noop_update_total": 8029,

        "is_throttled": false,

        "throttle_time": "0s",

        "throttle_time_in_millis": 0

    }

}

然后用[_stats?human=true]API查看索引信息，返回结果中_all中的total字段是所有索引信息的总和，结果如下：

"_all":     

    "total": {

        "indexing": {

            "index_total": 17039144612,

            "index_time": "12.6d",

            "index_time_in_millis": 1090693771,

            "index_current": 0,

            "index_failed": 9,

            "delete_total": 61,

            "delete_time": "62ms",

            "delete_time_in_millis": 62,

            "delete_current": 0,

            "noop_update_total": 24,

            "is_throttled": false,

            "throttle_time": "0s",

            "throttle_time_in_millis": 0

        }

    }

}

为什么会出现单个node的indexing指标会比整个集群的所有索引的indexing指标和的值还大的情况？
在做监控的时候，索引数量到底按哪个为准？