Nah, no shenanigans. It was definitely a glitch, but I'm not sure if I can explain it much better. The "flying cars" start with one wheel on top of the track, and one wheel below the track. As the wheels turn, it makes very little/slow progress, but you can see the car sort of shaking/struggling to release. That release eventually happens as the lower wheel gets to a point where it can cross back to the top of the track, and that twisting motion flings the car up and forward very quickly. The cars never even see the track again. I'm not sure what actually causes the car to stop... the replay wasn't very interesting to watch after the initial 'launch', as it just spins in whitespace for a little while and then eventually just stops spinning in an upright position while travelling forward.

I have a 200.24 d:196.51 h:-17.24/3.29m (gen 55).

At the 190 mark the world goes all glitchy and half the track disappears.

It's a shame it's not as deterministic as it could be (according to the blurb at the bottom of the page), since this [probably] means the genetic algorithm can't exploit quirks and things being "just right" optimally. Rather, it has to build something that is a bit more generalised.