In the join recipe, datasets can easily be swapped with the replace dataset feature. But in the stack recipe, the only way to replace datasets is to delete and add input. But this doesn't preserve manual column mappings. There are some cases when datasets need further preparation before a stack recipe, and so an analyst might need to add a prepare recipe before the stack recipe and replace the dataset input for the stack recipe with the newly-prepared dataset. This is pretty frequent when blending data from multiple sources. This feature would speed up that workflow and prevent disruption of mental flow when blending data.
