Designing for failure modes (bad network, old devices, no JS) often leads to better systems even in the happy path. This is a good case study of that.