Pairing the keyboard and the module is for both types of uses - for dumping the keyboard's identical programs onto the module, as well as for using different programs on each unit. In the first case, the point is to very easily (without programming) achieve a stereo depth by panning the two identical sounds at the mixer. This is essential for creating rich warm strings, brass, organ, pads, as well as monophonic patches. In the second case, it's for twelve-voice bi-timbrality or six-voice quad-timbrality. In both cases, it also allows for processing the layers differently. Another technique is to slightly alter some aspect of one program's settings. For example, setting the filter of one layer differently will create a stereo filter. The first sound in this demonstration is an example of that and it does not use delay ().
Folks have criticized this method as "overkill". Not at all. Personally, I don't like the sound of a mono signal. To my ears it sounds dull, lifeless, and very unnatural. My sonic standard with synthesizer is not the synthesizer, but the pipe organ, orchestra, and other acoustic ensembles or instruments. My goal is always to de-accentuate its electronic character and to make it sound more acoustic. So, with few exceptions, I maintain a stereo depth in my music and sounds. Whether or not it's obvious on YouTube is a different question, but the original recordings sound gorgeous for their deep stereo.