That's unbelievable! As in "I did not believe it". I mean it's nominally true, but only because this table has almost nothing in it, so a second index on random uuid is a visible cost. If this was actual table in actual software it would hold few other indexes and many fields and uuid+rowid vs int primary key would be a rounding error.

I never wished, gee, why didn't I use integer key. But so many times I wished I used uuid because eventually your data rows are going to need to have identity that is not local to this specific database instance.