The problem with smoke alarms is that it may take them months (if not years) to deplet a battery's charge to the point they won't work again. Unless you start with an already used battery, whose initial state is unkown, I don't consider them a good test.
They are, however, useful in proving that using bateroos in them might render them useless, thus endangering those who rely on such alarms.
Again, and sorry for insisting on this, a constant current and/or constant power circuit, like the one I provided a few pages back, might provide a better test, since it can drain a battery in a very short time, and you can still program at with rate the battery is depleted, thus emulating several real world scenarios with a single testing device.