When having a link to e.g. http://www.chumba.com/media/Chumbawamba-EnoughIsEnough.mp3 inside an anquor tag ikiwiki seems to parse EnoughIsEnough as WikiWord and breaks the link with that. As a general rule I would suggest that inside tags there should never be any WikiWord parsing, this is just asking for problems.
You can see an example of the breakage on http://wiki.debian-community.org/planets/de/ - scroll down to the Chumbawamba entry.
There's a great workaround for this bug: Disable the camelcase plugin. I really don't recommend using that plugin. Especially not when aggregating third-party content as you do in the example.
Fixing this at the html parsing level would involve making ikiwiki 2 times slower, not even counting the html parsing overhead, since it would have to fully render pages in the "scan" pass.
All I can do is improve the regexp it uses to try to avoid false positives. Which I've now done.