In the AWS context, they are, in fact. That's the original point of them - so during big AZ failures your reserved instances had first dibs on the available capacity.

The billing thing became more of the point as big AZ failures are so rare.