Lots of weird requests which resulted in 404s showed up in the log. Things like:
XX at http:/kaioa.com XX at XX XXathttp:/kaioa.com
Where "XX" stands for the node id. E.g. "36" for this blog post.
After investigating it for a bit I found the shocking reason behind this. Well, not really that shocking... it's more on the silly side, really. ;)
My RSS 2.0 feeds look like this:
[...] <link>http://kaioa.com/node/36</link> [...] <guid isPermaLink="false">36 at http://kaioa.com</guid> [...]
The RSS 2.0 feeds from Planet look like this, however:
[...] <guid>http://kaioa.com/36 at http://kaioa.com</guid> <link>http://kaioa.com/node/36</link> [...]
As you can see the isPermaLink attribute is missing. If it's missing it defaults to true, which in turn causes other readers/aggregators to treat that guid as URL. Ironically Planet does interpret that attribute for itself, but strips it from its own feeds.
isPermaLink="false" is used by Drupal and and WordPress. However, it only negatively affects Drupal's feeds, because WordPress' feeds happen to use guids which are identical to link. But that isn't a given and may change at some point in the future (well, it's unlikely).
Either way it's totally Planet's fault. I tried to track down the issue, but Planet's source is pretty hard to follow. Additionally "rss20.xml.tmpl" and the template stuff in general lack support for the isPermaLink attribute, which means that fixing it won't be that easy.
If you're wondering why I'm blogging about this instead of posting it on Planet's bug tracker... well, they don't have one. D'oh. I already contacted one of the authors, but so far I got no reply.