The clearest example I have is here: Simple tool for tiling any image to break up large engraving into smaller jobs
The raw SVG is provided in the linked post and discusses the issue (earlier in the post is a bit more discussion as well). There are 5 different versions of essentially the same image. If they were numbered 1-5 from left to right, the first, second and fourth are a 2 pixel x 2 pixel image with a simple black/white checker pattern. Each are embedded in the SVG with an image tag using a different image-rendering property (optimizeQuality, optimizeSpeed and pixelated). Glowforge treats them all the same and appears to use bilinear interpolation when sampling subpixel values.
The far right image is the same 2x2 pixel image, but represented with 4 separate 1 pixel images. You can see that the Glowforge engraves it very differently than the others due to the interpolation differences. This issue will affect any raster image that is broken up into tiles because the interpolation of the edge pixels on each tile will be engraved differently by the Glowforge. The linked case is particularly extreme because all 4 pixels are edge pixels and a single pixel is blown up to 1inch in size. In an image that is a normal size, this could present itself as a seam along the inner edges of the tiles. I don’t have a pass through slot, but if the process of splitting the image involves rasterizing into separate images, I imagine this would affect that as well.