Technically the problem was with the newsreaders (in Steven’s case Straw and in Dan’s case NewzCrawler), not the RSS feed itself. I turned on gzip compression on my Apache server about three weeks ago to preserve bandwidth. In theory, user-agents like web browsers and RSS readers should be able to either send a “Accept-Encoding: gzip” header and accept compressed data, or, if they don’t, be sent the original non-compressed data.
Somehow this mechanism was causing problems for Steven and Dan and their RSS readers (although it didn’t affect NetNewsWire, which is what I use). For the time being, I’ve turned off compression, and this appears to have solved the problem.