This is a great example of how nice it can be to reason about refraction stuff using Fermat's principle.
Let's reduce all this to 2 dimensions. The surface tension produces something like this:
Now if we want to know where a light “ray” needs to go to get from some light source, we just need to find the way that takes it the least time. Light is slower in water, so it wants to go as far as possible in air – of course, only if that's not too much longer to go. So far from the insect, a light ray would just enter the water straight perpendicular, since that minimises both the total waylength and the path in water.

However, right below the insect foot, that won't work – the foot itself is not translucent† – and, more importantly, a little way left or right from right below the foot the quickest path will still be right through the foot, since any other path will require the light to travel substantially more through water, while the total path length is only a little shorter.

so all these rays are “invisible”. Whether that works out this way depends on how far we are away from right-below-the-foot, so that makes a circular shadow, even when the foot itself has another shape.
†Actually, it is kind of translucent I suppose, but we know the small foot will only get hit by a tiny amount of light. So if that bit of light has to be spread over a whole lot of ground, there won't be much intensity down there.