Ok, At least two problems with this analysis.
1. How about heat from flexing that lowers the stiffness and allows more the top of the band to sag as speed increases. If this test was properly designed however, the sag from this effect would be countered by #2 below:
2. Doing this test in a drum means that there's no airflow over the top of the rubber band. If the band was rolling downhill at speed, the top of the rubber band would acquire an attached airflow creating lift due to the bernoulli effect. This would cause the top of the rubber band to bow upwards.
I'm amazed that no mention or consideration is made of either of these points.
It appears that PL/I (and its dialects) is, or will be, the most widely used higher level language for systems programming. -- J. Sammet