Looks like we've found your problem, then. The aperture should respond smoothly and consistently to the position of that linkage, regardless of the speed at which you move it. Some times, it's staying open when it should close down.
I just now downloaded one of your pictures, and examined the EXIF data to see how the lens is identified. It appears to identify as a G-type lens, which means no aperture ring, right? The sole control of the aperture should be via that linkage.
If it's not mounted on a camera, and nothing is touching that linage, then it should be spring-loaded so that the linkage moves downward, and the aperture closes to its smallest setting. If the aperture was wide open when you first looked at the lens, with it not mounted on the camera, and nothing pushing that linkage upward, then that's not right. The aperture should only open up when that linkage is pushed upward, and should immediately close back down when the linkage is released.
I see, also, that at 600mm, the maximum aperture is ƒ/6.3. That's only a stop up from the ƒ/9 that it's supposed to be at in your pictures; which seems consistent with the variance that we're seeing. The EXIF data don't seem to indicate what the minimum aperture is. I'd be curious to see this same issue manifested in pictures that were supposed to be at the minimum aperture.
In any event, it's obvious, now, what the problem is. Your aperture is sticky. You might be able to loosen it by repeatedly working that linkage manually, but you'll probably need to have the lens taken apart, and the aperture mechanisms cleaned and lubricated by someone who has the proper skills, equipment, and materials to do so. I'm just barely on the edge of thinking that might I dare attempt such a thing on my primitive non-AI lenses from the 1960s and 1970s. I understand these more modern lenses to be a whole other order of complexity, rather far beyond what any amateur ought to think of attempting.