Hmm. I'll have to think about that. (this is melting my brain!)
It's very difficult to think about - I constantly have to stop and think about which way things are moving, and usually with a "close one eye and move your head" test.
Three points I find helpful to keep in mind are:
1) The camera is always pointing at the object, because it's locked on to it.
2) Distant objects move WITH the camera (relative to closer objects)
3) Parallax is entirely from relative positions, not view angles (rotating the camera moves everything equally)
Unfortunate none of these are themselves super intuitive things either - best to try to set up a practical recreation of the scenario, or at the very least draw all the lines of sight in your diagrams.
#2 implies a static scene, but if an object is moving, then it's still useful, as the parallax (apparent) motion is added to actual motion. Thus a plane landing in the opposite direction to which you are driving can have it's motion cancelled out by parallax, relative to nearby trees.
In the context of Gimbal we see the clouds move right and we are looking (initially) left 54°. The clouds are moving WITH the camera which implies (point #2) the clouds are more distant than the object.
For the object to be far behind the clouds, it would have to be moving left fast enough to overcome the rightward parallax effect (which would be WITH the camera). A strong point against this is that as the camera's rightward motion decreases (as it turns to fly more directly towards the object), then the object's apparent motion relative to the cloud decreases, when parallax dictates it would increase.
I don't think this is possible with the Atlas hypothesis.
(Or course we also have the weird object scenario, where it's just flying over the clouds, then stopping and hovering - coincidentally as any parallax is small, but that's obviously not Atlas)