The proposed use cases in the paper like "forest monitoring," etc. seem like uses cases in highly dynamic environments. It seems like the applications will lend themselves more towards human pilots for the time being.
Fun experiment and test, though! Very interesting.
I'd be willing to bet they'll get it racing in the dark much better than humans can...
Only if it's using non-vision sensors. Plenty of FPV drones fly at night, and it's all down to pilot expertise.