You can eliminate impractical approaches all day long. At some point you have to start with the practical approach that you know will work a-priori.

When you have the option, sure! Not all problems are so tractable.

Granted. Still, I think there is a fundamental difference: In theoretical physics, a good theory of how things work (or how they definitely don't work) is enough as the result of your work, you don't need a working product in the end. In practical engineering, you do!

Let me also add that I am serious about theoretical engineering, actually that's one of my favourite occupations, together with practical engineering.