It does not necessary work as you would think as the model is very vague and this causes GPS elevation to be very vague
I already did the math, on the Pikes Peak ascent where you climb 7,815 ft in a half marathon distance your average grade is 11% which will result in roughly 5% distance error. If you're "running" a race that is steeper than this you're probably hiking and not running. And we're still talking about only 5% error in this extreme scenario. I really don't know a lot of runners who can run up anything steeper than an average 10% grade and still call it running. To have a 10% error in distance you would have to average a 14.2% grade for the entire race. I dare you to find me a foot race with 14% average grade anywhere in the world then I dare you to actually "run" it.
So the vertical does add up but it's still marginal to the error you get from GPS inaccuracies on the trail with the Fenix 3.
Cycling is probably the biggest question mark. You can handle a bit steeper of grades on a bike, although still probably not steep enough to be worth it on 90% of rides... but if you have a speed sensor on your wheel, then the speed you get from that is effectively 3D. So would you not want to have 3D distance on too?