The fix will be available in the next official release Q1 2015. 

RadRichTextBox's behavior is to reject/accept the whole revision where the caret or selection is. Modify the behavior so only the selected region is rejected/accepted and the rest is split into two revisions. Consider multiple selection ranges as well.

When ignoring incorrect words through the Spelling dialog and you reach the last word in the document which is incorrect, only one more word from the beginning of the document is checked prior showing a message that the check is complete. Instead, all incorrect words starting from the beginning of the document should be spellchecked. 

The issue is reproduced when adding the incorrect words to the dictionary. In order to reproduce the issue, only "Ignore" or "Add to Dictionary" action should be applied. When the actions are mixed, the issue is not reproduced.

Workaround: before opening the spellchecking dialog, move caret at the beginning of the document
this.radRichTextBox.Document.CaretPosition.MoveToFirstPositionInDocument();

IndexOutOfRangeException is thrown when some JPEG file in the document has Scan Header with fewer components than the Frame Header. The current JpegDecoder implementation requires both JPEG Headers to have the same number of components.

WORKAROUND: PdfImagesCompressionMode.Automatic option may be used in the PdfFormatProvider export settings. This way all JPEG images will be exported to PDF without decoding their pixels and the PDF file will be exported successfully. 

As a result RadDocument instances cannot be collected by the GC.

ContextMenu holds a reference to last RadRichTextBox it's shown for. Since ContextMenu is cached by MEF, this leads to memory leak. 

Workaround: There are two options:

- manually set different instances of ContextMenu to all your RadRichTextBox-es
this.radRichTextBox.ContextMenu = new Telerik.Windows.Controls.RichTextBoxUI.ContextMenu();

- Create custom context menu that will be non-sharable. MEF will automatically load different instances of it for all RadRichTextBoxes on demand:
[PartCreationPolicy(CreationPolicy.NonShared)]
[CustomContextMenu]
public class NonSharedContextMenu : ContextMenu
{
}

Using any of the options will prevent MEF from creating cached instance of the ContextMenu.

When a document with NumPages field in header or footer is exported to PDF, NumPages field is evaluated wrongly - 1 for the first page, 2 for the second page, etc.

The bug is regression, introduced with R3 2017 SP1 (2017_3_1018).

Fix available in LIB Version 2017.3.1211.

If the empty run is defined with start-tag and end-tag the import is correct:

<w:p>
  <w:r></w:r>
</w:p>

However, if it is defined with empty-element tag, its parent Paragraph is not imported:

<w:p>
 <w:r/>
</w:p>

Fix available in R2 2018 SP1 release version.

When the document uses a font that cannot be found on the machine, it is substituted for another font. What should happen is to fall back to another font for the visualization, but to preserve the original font in the model.

Partial workaround: for HTML import, the font can be preserved by subscribing to the FontSubstituting event:
provider.ImportSettings.FontSubstituting += ImportSettings_FontSubstituting;
private void ImportSettings_FontSubstituting(object sender, FontSubstitutingEventArgs e)
{
      e.SubstitutionFontFamily = new FontFamily(e.OriginalFontName);
}

When a merged cell contains more than one entire column, the caret is in the merged cell and you chose the option to delete the current column, the action is supposed to delete all the columns in the merged cell. What happens instead is that only the first column is deleted and the merged cell remains.

Workaround:

            TableCell tableCell = this.radRichTextBox.Document.CaretPosition.GetCurrentInline().Parent.Parent as TableCell;

            if(tableCell == null)
            {
                return;
            }

            int columnSpan = tableCell.ColumnSpan;

            this.radRichTextBox.Document.History.BeginUndoGroup();

            for (int i = 0; i < columnSpan; i++)
            {
                this.radRichTextBox.DeleteTableColumn();
            }

            this.radRichTextBox.Document.History.EndUndoGroup("Delete Column");

When a table is selected, the SelectionMiniToolbar should have a borders drop down menu which allows setting and editing of borders without entering the table borders dialog.

WMF contains both bitmap and vector elements and currently RTB supports only the bitmap parts. We should consider adding support for the vector elements as well. 

Hyphenated words (like "good-looking", "dry-clean", "twenty-seven") should be passed to the spellchecker as a whole word. The current mechanism works well enough for English, however, for some languages and dictionaries it is necessary to pass the whole hyphenated word to the spellchecker.

Create document with some text.
Insert image.
Right click on image and add hyperlink to it.
Drag the image somewhere in the text.
Observe that it is created a copy of the image. 

Actually, the issue occurs when the hyperlink is inserted.

If nested comments are present in the document, the end users cannot delete the nested one. If they put the caret inside the inner comment, or even inside the balloon associated with the inner comment, and press "Delete Comment", the most outermost comment is deleted instead of the current one.

Current editing style properties are not modified by Font dialog. Also applying through methods of RadRichTextBox or RadDocumentEditor does not work.

When a table cell contains nested table, and the paragraph after the nested table is empty, this paragraph could be collapsed, as in MS Word. This gives better look to nested tables, similar to how browsers visualize nested HTML tables.

Note that the paragraph should be visualized when the caret moves the position inside the paragraph.

Currently the paragraph is visualized with its full height, making the layout different than MS Word - it appears that there are redundant space after the nested table.

Adding a custom list could be done by using the AddCustomListStyle method of RadDocument. However, this operation will not be included in the Undo stack.

Deleting the full chain of relationships between a list style and a document list is not possible.

Instances of RadDocument are leaking in addition to other objects when the CreateDeepCopy method of a Section is called.
The section contains one paragraph in its blocks and has one paragraph in the default header.

Fix available in LIB Version 2017.3.1002.