tl;dr - As the title states - If I disable a non-clustered index on a table, are the linked statistics for that index still used?
I'm aware there are many questions about "dropping vs disabling indexes"...but none that I could find that specifically cover statistics.
I know that the statistics are not altered or changed (at least that's what I gathered from the MS documentation). But my question is whether the statistics are still used.
For background, I'm working on a large index tuning project. It involves adding/removing indexes across hundreds of identical databases but with varying workload patterns. Collectively there's over 2M indexes.
My first step is to drop any "unused" indexes. However, rather than dropping them, I'm considering just disabling them instead so that the definition is retained. This would allow me to record in a table the instance, database, object id and index name/id of any index that is disabled. If performance begins to degrade after, the index can be re-enabled (rebuilt).
However, if the statistics for the disabled index are still used to generate plans...then disabling them will not have the same performance impact as dropping them. If this is the case, then disabling the index is not a "true" performance impact test and I risk introducing performance problems if the disabled indexes are eventually dropped.