The markdown processor and @image tag seem to be interacting in a strange way which triggers unintended code sections.  In the below example (from doxygen-users "C# comments and blank lines"):

        /// <summary>
        ///             Brief text.
        /// </summary>
        /// <remarks>
        ///             Text.
        ///             @image html picture.png
        ///
        ///             More text.
        /// </remarks>
        class Test {
        }

The section "More text" is taken to be a code block, even though it is not indented from previous comments >3 spaces.  One workaround noted was to replace the blank line with a manual <br>, at which point doxygen would no longer trigger the code block as the line is not preceded by a blank line.

Another example (from doxygen-users "Markdown: images and paragraph indentations") is shown below:

      /**
      \page newpage Some Test Page
       text

        text

         text

          Start talking about the image

          \image latex someimage.png "An image"

          And continue talking about that subject
      */

In this case, the "And continue talking" is interpreted as a code section, again even though it is not indented >3 spaces from the previous comment line.

I'm suspicious that the @image tag is reverting the parser's internal knowledge of indentation levels--such as causing doxygen to think that that line is not indented.  Then when it comes across the "And continue talking" line, it finds it indented >3 spaces from the previous line and treats it as a code section.

The preferred behavior from my perspective would be to not trigger the code sections in these cases.