As you must know, a Kettle transformation is a group of linked steps through which data flows. Each step is meant to receive rows of data, process the data somehow, and deliver those rows to the next step or steps. If you agree with this definition, then you must realize that if there are no rows coming to the step, the step will not be executed.
This seems reasonable, but on occasion, it can be a problem. To get an idea of that kind of situation, look at the following scenarios: