Isn't that still prescriptivism? Descriptivism here would mean noting what people claim constitutes sandwichhood, and maybe noting what people's behavior reveals about what they actually do or don't treat as a sandwich. But it would be prescriptivist to assert a new definition of sandwichhood, even if you argue that it's somehow more motivated than others.
A prescriptivist says that a pronoun should match the number, gender, and human-ness of its antecedent, and therefore “they” is not acceptable when referring to a single person. A casual descriptivist says that they understand what it means when other people use “they” in a singular fashion, and find that it is a useful construct. A formal descriptivist would try to describe how the usage and clarity is practically evolving over time. The terms “pronoun” and “antecedent” are terms themselves modeled from actual usage, rather than some organizing committee that decided how English should work.
Similarly for sandwich definition. The descriptivist viewpoint says that no strict definition is necessary; that people have a casual understanding and should model their usage on that. A prescriptivist says “here is a definition of ‘sandwich’ that models a significant number of things that many people agree are sandwiches, so we expand that definition and use that to label things as sandwich or not sandwich, regardless of the common understanding”.
The key point here is that I do not feel a strict definition is necessary in order to reasonably discuss sandwich labeling.
Practically speaking, I have found that my heuristic above generally yields agreement from others (in particular, I have not yet found anyone who would be happy with a hot dog under those circumstances), with occasional differences of opinion that would be interesting for a professional sandwichologist to examine as a question of sandwichial drift over time.
Or put differently: if reality and models differ, a descriptivist says that the models are wrong, while a prescriptivist says that reality is wrong.