Hey @Joel_Baerg,
Try this workaround: first extract URLs of those images, cross-doc the URLs (could be an issue of its own, you may want to .ToText() them first), then turn back into images with Image() formula (most likely in a FormulaMap()) on the receiving side.
If the endpoint that serves images is dumb enough (i.e. doesn’t check what page tries to load the images, the original one or some other doc), it should work. I have a feeling it should.
Also re 125 MB, it’s only for the doc, not the images (most likely, as they are stored separately from JSONs). It’s quite a lot of headroom actually. And also this limit only matters for cross-doc — opening the doc would still work.