I've questioned maybe 200 people about this over the years and when someone starts talking about spatial understanding and not seeing pixels, they aren't really talking about seeing anything in mind, but more understanding it. I can do the "feeling" thing, which I reference as "I have the mesh, not the map". People with Aphantasia appear to hold facts about objects, but not actually generate the imagery where they are conscious of seeing it. It would be a little like having Dalle3 generate an image from a Claude Opus 3 prompt, then uploading it to Claude to look at. It can't do generative images, but can look at them and inference.
Maybe someone that visualizes strongly here can confirm that what you are indeed seeing has aspects of light, shadows, color and the things we consider attributes of "pixels"?
I wonder if being good at a specific internal style (e.g. visual, verbal, etc.) translates to being good at specific types of problems, or is the core capability an abstract foundational capability and the different styles are really just a method of presentation.
My imaginary scene clearly had some "spatial sense" though — I saw (but more like "felt") the flat surface of the table, the edges of it, how it is positioned relative to myself, the roundness of the ball rolling, and how it falls off.