Yes I would think so too - I just haven't been able to make it behave the way we would expect yet.

It's currently sat in my interesting-but-useless pile. The only thing that interests me is how the server enforces the repeatability of the data set, and whether the mechanism used somehow also guarantees order. For example, does it keep a list of page IDs somewhere and always scan them in the same order, or does it make a temporary copy of the data?

The ordered:false in the query plan is relevant; I just wonder whether the storage engine does anything special for tablesample...

Paul