From Matthew Thomas’ post “Why Free Software usability tends to suck”:
“The practice of releasing early, releasing often frequently causes severe damage to the interface. When a feature is incomplete, buggy, or slow, people get used to the incompleteness, or introduce preferences to cope with the bugginess or slowness. Then when the feature is finished, people complain about the completeness or try to retain the preferences. Similarly, when something has an inefficient design, people get used to the inefficiency, and complain when it becomes efficient. As a result, more user preferences get added, making the interface worse.”
[Found via Phil]
Petafloptimism 
I think the reason really is that most of the OSS developers are just really bad at saying “NO!”.
Especially if your motivation is to gain reputation and get bananas and milkshakes from your users, not money; then it’s often really hard to say no to people who flatter you first by saying how much they like your software, and then asking for something.
And thus features and preferences get added.
It’s also that broken usability doesn’t actually stop anyone from using the software. So it seems more logical to add a new feature.
And most programmers think usability is something done by dotcom graphic designer dudes and thus not their business.