It could cause distortion in both pictures, one, or none depending on the angle of the camera, height, distance relative to the window. I am not saying that I am concluding that this is the case, just that we do not know if it is and therefore cannot draw absolute conclusions. I was referring to the distance of the visible horizon.
The short answer I do not think it is realistic to still dispute the pictures, because it cannot be
absolute conclusive. However, then you can doubt everything, because you will never get absolute conclusive evidence in anything. If you would run a laboratory test in which you get perfect conditions, there are still factor which change. If you take a camera, press once, press twice. The second picture was taken on a camera with slightly more wear and tear, the battery was not as full as when the first picture was taken, the time at which it was taken is different. The force at which the laborant pushed down on the button. Etc. All small changing factors.
All these small factors are neglectable, unless you got sufficient reason to think otherwise.
The longer answer Nothing in life can be completely certain. However, looking at the things we observe. In the pictures you have presented we can clearly identify the reasons behind the curvature.
We have looked at the different photos and could see that if the horizon was captured high on the picture it shows a downward curvature. If the horizon was captured low on the picture it shows an upward curavture. We also looked a picture in which the horizon was captured almost in the middle. The horizon was captured slightly below the middle, so therefore we can notice a slightly upward curvature, but this one is not as significant as the lower one. Even in these pictures some small factors may be in play, the angle of the camera, the distance of the photographer to the sea, the zoom, but none of these have such a factor that they change our set of rules.
The rules are:
If the horizon is high on the picture, it curves downwards, due to the wide angle lens.
If the horizon is captured in the middle, it is horizontal, due to the wide angle lens.
If the horizon is low on the picture it curves upwards, due to the wide angle lens.
You can rerun this to verify the rules by using another wide angle lens. You could also verify it by
not using a wide angle lens. If you use a normal lens and you capture the horizon high on the picture you should
not see a downward curvature, etc. It is about ruling out the variables so you are left with just one adjustable variable.
In the pictures I have presented to you. If any small factor would be in play, what would cause such a major distortion? Can you rule out anything in particular? To me, the size of the engine is pretty much the same, the height at which the horizon was captured on the picture, the perspective are also almost identical in both pictures. The only
huge difference between two pictures is the altitude. So the set rules we can apply here is.
At low altitude you cannot detect a (significant) curvature
At high altitude you can detect a (significant) curvature
And just to verify if these set of rules are true, you can rerun it. Go in a plane, take it to a high altitude and see if the horizon will curve downward. Also if you stand on the beach at low altitude, does it curve? It will give you conclusive evidence the set rules are true. Any small factors are neglectable, unless you got sufficient reason to believe otherwise.