Have you thought about some type of preprocessing to make the PDFs "simpler"? https://github.com/freedomofpress/dangerzone does something like that in its first stage.

While we don't use this yet, it seems very promising - thanks! We did something similar (with libreoffice, for example) to have support for non PDF datatypes, but this seems like it is coming at it more from a security perspective - which makes sense.