The update r133's image deduplication got me thinking: wouldn't it make sense to separate textures from game objects (sprite, tiled background etc.)?
Essentially, a texture object would only contain image data. Sprites, tiled backgrounds would then refer to those texture objects to pull images.
This way, you could manage all core image functionality in one place (think spritesheeting, compression/decompression, loading etc.), whereas sprites and other objects would only need to deal with positioning, resizing, tiling, z-ordering etc.
Just a thought. I am aware that I have little idea of how deeply wired Construct 2 is and that redesigning may already be too big of a task at this point, but it might actually make things easier in the future, who knows?