I started moving to MVVM as the project progressed and I began to see the limitations / issues with using UserControls and code-behind.
To be honest the application grew from just hacking things together initially so there are some older parts that were just me picking up the technology.
As time progresses I'm trying to move more and more of it to MVVM and get everything covered by unit tests - at the moment the majority of unit test coverage is on the server side - the client is a little light in that department atm.
Word document conversion is a nightmare, and is some of the more obtuse code I've ever written. At the moment we convert to XPS and work from there but I'm sorely tempted to go to images and not loose too much sleep over the fact that you then get pixellation as you zoom in a lot.
I'm curious about the architecture. Did you use Model View ViewModel?
How are you converting the word docs? Do you convert every type of resource to an image?