Yes, I've noted this too, and as Mr. Hound reported, the whether any single row is updated or not is irrelevant, it's more to do with if any row on a specific page is changed that decides if it gets written to disk - that is, assuming we are correct in thinking that SQLite checks it so. The non changes may be brought about by a completely other mechanism. Either way, I'm thinking more and more that SQLite writes only when needed and any query parts to try and "help" it is a waste of CPU cycles - but I am still not 100% sure.