That sounds like a nice idea, but though the filters are roughly static, their behavior changes quite a bit for the data. For one run, I might want to exclude all objects of property X, with a value of Y below a particular threshold and which wasn't previously selected in the past hour. For the next run, X might be irrelevant (though the filter still gets run because we don't know that until the filter is run), Y might have a different threshold and we might not care about when the object was selected. So multiple filter runs with the same set of objects have, literally, millions of possible different outcomes.