So gave it a go and added in fractal noise based displacement maps and compositing -nothing groundbreaking from the point of the previous filters, just fine-tuning the filtering process upon the experiences.
These are in general: never use the background image as an input for the filter chain. Forget setting an object into overlay blending mode atop another. It's worthless.
Either pull in the affected objects with the image filter primitive -which takes other serious precautions and setting up the image with the right origins and whatnot- or, apply one filter after another to the "same object".
Surely you lose the option to use separate fill attributes or shapes but actually get to an image which will more likely be rendered.
All in all: comlex filter chain setups need to be broken down to subchains, and, maybe the next filter editor will be able stitching them or the functionality will be adressed by an extension.
Grouping the objects with every "small" filter is a burden that cannot be easily tracked either.
svg source
