We're seeing similar requests at Lancaster, Ghostscript is used to convert the PDF files into PNGs which are then displayed on the grading screen
We have a request to make links clickable
A request to make the highlighter tool confirm to line height (it often ends up highlighting between lines)
A request to make the interface more accessible - screen readers can't read text in a PNG image
None of which seem doable whilst we have an image file to annotate. Other services, such as box annotate convert documents into standard HTML5 as opposed to images which works much better, IMHO
As you say, tutors can open the original file but that destroys the workflow.