I think I could word things a bit better: toolpaths are generated at the canvas' resolution, because that's the resolution that layers are composited at (i.e. the 'camera snapshot' analogy). A high resolution raster-layer is going to appear low resolution when the canvas resolution is low, and generate coarse toolpaths. Increasing the canvas resolution will allow more of the detail in a raster-layer's contents to come through, and be captured and represented when toolpathing. The raster-layer's contents and resolution are not changed by the canvas resolution though - they just might not be captured/represented if the canvas resolution is too low.
Another analogy: an old low-resolution webcam taking a photo of a painting doesn't change the contents of the painting, it just doesn't capture the details of the painting. Raster-layers are just paintings being placed around the canvas, and to generate a toolpath (and render it in the 3D view) the canvas is composited into a 'snapshot', at the canvas resolution.
The contents of a raster-layers does not change with canvas resolution changes, only the resolution that they're composited with the canvas at, which only affects rendering the canvas in the 3D view, and toolpaths generated from the canvas.
Ok, I hope my explanation wasn't annoying - I just felt like it was easy to interpret my previous reply the wrong way, almost diametrically opposite of what I was hoping it would convey :P
Let me know if you got what you needed :)
- Charlie