Macros aren't currently able to generate RSS with page contents. This is because the Markdown syntax is processed after the macros are executed.
I propose a patch which processes the contents after building the list of pages, but before calling the macros. This way when a page is being transformed to HTML, we already have the list of pages (in case some inline python code needs it). The HTML code is stored in the page list, so when the once_ macros are called afterwards, they have access to it.