I suspect that the time and cost involved in non-trivial structures will limit any real experimentation with non-traditional designs. It might be trivial to throw away a kilogram of plastic on a failed prototype, but the same doesn't extrapolate up to experiments with bridges.
But that's just the thing, you aren't limited so much by time and cost. The time is proportional to the size of what you're printing, and the cost to the amount of materials and time involved. None of which says much about what the design must look like. You design it on the computer, look at in 3D on the screen as much as you like, and print a scaled model, if you must.
You don't need to print up a full-scale prototype anymore than you would do that with a traditional design and construction technique.
I assume the printing of this bridge didn't go smoothly in one pass, but one advantage of larger-scale metal printing is if something goes wrong you can grind off the failure, reposition the print head (extruder? welder???), and try again.
I suppose you could do that with FDM as well, but the precision required for smaller prints is much greater.
