It talks about descriptive adequacy of a thought experiment, while the goal of the enterprise is to figure out what should be done.

I partially agree. But the point is, in any emergency situation, you're going to default to your training if you're acting. Thus, individual moral intuitions give way a host of other concerns, and a body of history, literature, and tradition of the particular discipline (whether it be emergency first response, engineering, soldiering, policing, surgery, or any other form of life or death issue).

If you're going to spend the thought cycles, much better to use a real discipline. Here's one - there's two run down apartment buildings with roughly 200 people in them. Mortars were fired off the rooftops the night before, killing ~20 innocent civilians. The next day, military troops raid the buildings, arrest everyone, find a cache of weapons, and strongly suspect the people using them are among the 200 arrested. Everyone says they don't know who did it. What do you do with those people?

It addresses the same questions a trolley problem does, except it doesn't have the flaws a trolley problem has.