Hi Krzysztof,
Very cool video, thanks for sharing!
I once read that glass surfaces can be hard to detect for bats an acute angles, I presume the same might be true for an even water surface.
It's interesting that two passes seem to be the norm for drinking, one approach pass and one return pass. Maybe the first pass is for obstacle clearance, and on the second (drinking) pass the ultrasonic beam is diverted downwards for better surface detection, flying the previously determined obstacle-free path for safety?
Not much time to watch the video as intensely as it would deserve, so I'm just quickly speculating here
Regards,
Henning