It's a pretty specific use, but it can come in handy.
Imagine you have a different values to stack, some data varying a lot and some being almost constant. If you use the default order and the variable data is stacked under the constant data, the variable data will make the constant data have a very variable base. So if you stack first at bottom the less variable data, it could help.
Example: These two graphs show how to improve readibility by stacking deeper data that move the less, i.e. has the smaller standard deviation.
Default graphing could lead to bad readibility
Improved readibility when sorting by standard deviation