Panning and zooming on very large images

Hello all,

I have started to learn Resolve and Fusion and so far I like what I can achieve

For a private project I want to do a travel route animation on a map. With two different compositions I want to achieve two different goals:

First: I show a static overview map of the area to be explored and animate the travel route taken with a MaskPaint-Node applied to a Background-Node merged over the image of the map itself. The WriteOn of the MaskPaint is keyframed, so I can animate a line snaking on the map. This works like a charm, no problem with that. I want to use this animation as an overview of the whole area/country whatever.

Second: I want to replicate the animation from my first goal, but on a much larger scale/image. While the map from the first animation is nice to give an overview, transitioning from one point of interest to the next happens on a much smaller scale/distance. So I need to zoom in quite a bit to a) show actual movement of the track on a map and b) show meaningful detail (labeling individual sites of an area in a view of the whole country may be sufficient for the Vatican or Andorra, but for a country like Iceland will result in all labels starting at the same pixel).

To achieve this second animation I would normally just take a large enough image (I think the original image of the map was around 13000x6500 px) and use this in fusion and just do some transforms as to move the "camera" around, but alas, 4GB of dedicated GPU memory is not enough.
What I have done next is: Take the image (and rescale it to a proper, divisible by 1080, 16:9 ratio) and slice this up into multiple tiles. I thus have ended up with 9 tiles of the map, each tile 3840x2160. I am not sure yet if I want to finally export to 4K or just HD, but having a rather large 4K display to view the result on, 4K seems the obvious choice.
Because I want to play with tilting the camera I have gone 3D-Nodes and the result looks like this:


Each image is merged under a MaskPainted Background-Node and piped into an ImagePlane3D. Each ImagePlane3D is offset by the correct dimensions to create a full image from the individual tiles. The rest is just merging everything (adding a Camera3D and Render3D) to generate an output.

This works, but:

• If the drawn route crosses the boundary between two tiles I have to fiddle with each MaskPaints WriteOn to properly keyframe it, so the animation of the path seems a) seamless and b) does not speed up or slow down depending on the path length on each individual tile
• I want to animate the full route in one composition, so that I can later put this composition (probably as a compound clip?) in my final timeline and trim it individually to the right portions of the route I want to show, so as not to recreate the composition for each and every transition part

So my question is:
Is this the least false way to do this, or is there a simpler way?

Something really fancy would be if I can tell fusion: This is the image (in the timeline resolution for example). And this is the exact same image(content) in 4 tiles. And in 9 tiles. Jump to the next set of tiles if you are zoomed too close. This would give me the opportunity to draw the route and place markers for points of interest once and the rest is done by fusion.
(I might have a go at this by programming it myself, but that would be outside of Resolve, so I would have to reimport an image sequence every time I change something with the length of the animation)