As for reasons / performance, library author would know best. I dont think there should be a reason instantiating the DOM node there is less performant - the library just would check the dynamic type and do it later with the array syntax form. On second thought, there may be some performance layer that runs faster with the array syntax
I was somewhat confused by the way it was phrased, but it seems like just a performance related question.
Why ever use the array syntax? Just convenience when it suffices? Is there any performance benefit to using it over instantiating a DOM node in application code - toDOM? Are all conformations going to operate predominately the same, performance wise?