As one of the people who was able to correctly tell which was PWM, I think there’s a few things to note - some of them are general remarks, some are comments about the quality of the test or the poll’s evaluation of the test’s quality. In no particular order,
The pwm here is of the faster sort that is not commonly a big problem. Yet close examination does reveal its presence, so it’s at least shown viable even for high frequencies of pwm which may be hard to notice.
The stream of water is one which we could not see with our own eyes, nor in motion or from multiple perspectives, and so we must guess at the speed and character of the water flow. This is true to an extent even in person, since most showerheads do not produce variation-free patterns of water.
It would have made me more sure of what I was seeing if I had a better idea of the specific camera settings and attributes, rather than just knowing the exposures were equal. Not exactly necessary, but while I may reasonably suppose that the iso was higher than the camera is able to be noise-free at, the aperture was opened wide enough that on whatever format this is, at whatever distance this is, the focal plane was a certain depth, and I may assume the exposure time was within a certain range, at least for me knowing as much as possible about the camera’s configuration helps me to subtract out that influence. We did know it was 30 lumens, which was useful to know. All this is partly because you can choose odd settings and get odd results, or choose settings that don’t make things very clear.
The picture quality is deceptive; you don’t realize until you zoom in that there is any more detail to be had - while I’m for the above reasons unable to be certain to the degree I would prefer, I would point to color noise, noise reduction size (likely basic built-in NR rather than purposefully applied) and large areas out of focus as reasons why someone not looking closely wouldn’t realize that within the full-size image are some smaller areas which are in good enough detail to see the fast pwm we’re looking for.
While 50/60Hz or even 100/120Hz flicker (not PWM) of some mains powered lighting allows things to travel fairly large distances between each cycle, the kilohertz scale pwm of a decent flashlight requires you to consider how far something can travel in a fraction of a millisecond, and then zooming in to look at that scale. The larger droplets smooth out the fast pwm so that it can’t be seen, we see only the effect of their changing shape. It is, however, conceivable to me that one might see similar patterns and correctly guess that there was a slower flicker or pwm going on. That gets into where this test isn’t very good; without great confidence in your analysis, you can come up with multiple plausible explanations for what you see, and be unable to choose.