Documentation on MDS : weird statement about write completion correlated to index updates

While scaling a distributed architecture is great for data, it is a challenge for indexes and queries. Simply put, scaling data speeds up reads and writes but it slows down queries, and while creating indexes speeds up queries it slows down writes. For example, if you scale out your database to support a massive amount of data, your query performance will suffer dramatically because the more nodes participating in a distributed query, the worse it performs. And creating many indexes to improve query performance will negatively impact write performance because writes do not complete until all indexes have been updated. The bottom line is that scaling out, with every database service running on every node, can result in a big drop in efficiency and performance. 

And creating many indexes to improve query performance will negatively impact write performance because writes do not complete until all indexes have been updated.