I think that part of the article is misleading and causing confusion; statistics never became out-of-date unless changes (inserts/updates/deletes) happen without updating statistics; that happens when statistic auto-updates are turned off, changes happen and statistics are not updated, but that is case where async updates can do more harm than good: SELECT queries are supposed to be the most consistent in terms of performance, if you have random stat updates in SELECT queries this consistency is broken. Just imagine you run a small query on a 1B rows in 20ms then all of the sudden a stat update will make your query execute in 5 minutes: shoot that DBA with salt, multiple times.