We have a good news for all of you who have experienced problems with the memory consumption of RadPdfViewer. We have made significant optimization of what is cached for each page when scrolling in a document.
Please have in mind that when using big images in a document the optimization may seem not so significant. This is because there is additional opportunity to further optimization of the caching mechanism for the image sources.
I have logged this additional opportunity as a different feature request. Here is a link if you would like to subscribe for status updates of that item to: Optimize the image source cache - http://feedback.telerik.com/Project/143/Feedback/Details/189305-optimize-the-image-source-cache

When using CTRL+Mousewheel to zoom in to a PDFViewer document, it should zoom into the point where the mouse pointer is located. Currently it just does a simple zoom of the whole document.

For example see Acrobat Reader or Google Maps.

Represents Highlights PDF content. While the other annotations are drawn on top of the content, the highlight annotation is behind the text it highlights. In other words, it has to be inserted at a specific place in the ContentElementsUILayer. The precise place where the highlight should be placed is described in the PDF file with the BDC, MCID and EMC operators. At the moment, we have implemented a read-only rendering for all annotation types which draws the annotions separately from the page content. In the case of Highlight, the annotation is drawn below the other content. With this implementation, depending on the PDF content, the highlight annotation might be visible or it might not be because in some cases there is geometry content below the highlighted text which hides the highlight annotation appearance. 

When a PDF document contains images compressed with CCITTFaxDecode with applied BlackIs1 parameter, black color is visualized as white and white as black.

A rubber stamp annotation displays text or graphics intended to look as if they were stamped on the page with a rubber stamp. When opened, it displays a pop-up window containing the text of the associated note.

Layers represent groups of optional content which can be hidden or shown.

Most PDF viewers have option (enabled by default) which allows fields to be highlighted with colors predefined in the settings. Required fields are highlighted using different color.

For example: In Adobe Acrobat Reader DC this is enabled in "Edit" -> "Preferences" -> "Forms" -> "Highlight Color" -> "Show border hover color for fields". 

This would allow switching the focus between form field controls by using the Tab key.

Scanned document is not rendered with the new engine

When displaying PDF-Files using PDFViewer, ContentElementsCanvas .RenderAsync uses reflection (DispatcherObjectUtils.ApplyDispatcher) to render Visuals on multiple threads.
However this leads to a memory leak since the dispatchers cant be GC'ed, see screenshot below.




The Screenshot is from the actual application we expirienced this issue with, the attached reproduction example is a boiled down version of what your code does.
If you wanna reproduce this on your own, create an application that uses PdfViewer that switches between many pdf files. The ammount of Dispatchers will grow steadily, probably to a total of the number of threads used by Task.Factory.

As a <role> I want <ability> so that <benefit>.

Printing document on a machine with .NET Framework 4.6 installed does not work for some documents.

Currently, an issue is opened at Microsoft Connect: "Printing with PrintDialog fails when .NET 4.6 is installed" (https://connect.microsoft.com/VisualStudio/feedback/details/1980419/printing-with-printdialog-fails-when-net-4-6-is-installed).

Workaround: In the attached project.

Add support for importing and filling interactive forms.

Interactive Forms content is visualized read-only in RadPdfViewer UI with R1 2017 Release

Available in R3 2017 Official Release Version.

When users do "copy" operation, add the content to the clipboard in RTF format. This would allow to paste the text along with its formatting in text editors like MS Word.

The images are loaded asynchronously and when scrolling too fast, the loader might fail to load an image if it is bigger.

This is caused by an incorrect order of applying the TextMatrix and the glyph stroking operation.

Visualizing a page with a huge number of images (40000) freezes the application.
When switching to the FixedDocumentSinglePageViewPresenter, all the page content should be loaded so it can be later rendered. Although the images are loaded asynchronously, there are too many threads started which seems to cause the hang. Also, according to profiling results, it seems like the image caching causes performance diminishments in such scenarios. 

Provide a functionality to show one page at a time mode.