For a little while I was pretty intent on the object in Gimbal being Venus -- its angular diameter is in the neighborhood of 50 arc-seconds, and Venus' angular diameter varies between 10 and 66 arc-seconds. You can also find some amateur IR photos of Venus that include significant glare.
I've changed my mind, however. The object image gets 20.1% larger in area over the course of the clip. This seems to falsify the Venus hypothesis.
To find an empirical measure, I made four screenshots from the NYT version of Gimbal -- one at the first frame (in which the object is shot in white-hot), one from the last white-hot frame, one from the first black-hot frame, and the last frame. In Photoshop I used the magic eraser tool with a tolerance of 64 to isolate the object in each screenshot, then cranked up the contrast, then counted the white or black pixels. Findings:
1. The image in the white-hot segment increases in area from 2507 to 2836 pixels.
2. The image in the differently imaged black-hot segment increases in area from 4115 pixels to 4367 pixels.
3. To make the white-hot segment comparable to the black-hot segment, I multiplied the white-hot measurements by a factor of 1.45. After this adjustment, the white-hot segment increases in effective area from 3635 pixels to 4112 pixels.
4. Thus the effective area increase of the image, over the course of the entire video, is from 3635 pixels to 4367 pixels, an increase in area of 20.1%.
This suggests that the object gets about 9% closer over the course of the video.
The horizon issue mentioned above is a problem, too.