We are passing a list of dictionaries with verbose_name and identifier, we should use a data structure for this instead to avoid bugs.
We need to make sure that celery can serialize named tuples.

ref readthedocs/readthedocs.org#7548 (comment)

You can have two API objects that use the same anchor reference. When both are rendered into the same Markdown file, the references are not unique, and one of the links will not work.

def a():
  """ Links to [a][0].

  [0]: https://a.org """

def b():
  """ Links to [b][0].

  [0]: https://b.org """

The MarkdownRenderer should find such duplicate references (here 0 an

It would be great if there was a function in the theme handler in which I could modify the HTML before it's converted to PDF, for example to add a title page.

(At the moment, I tried to add a title page by combining the output with a pre-made PDF for the title page using Ghostscript, but the problem is that Ghostscript ruins all the named destinations links, so internal links won't work. Would