Before it generates the OPML, it makes a series of calls through the Twitter API to find out who the user is subscribed to. Twitter only returns 100 names at a time, so if the user subscribes to 1000 people, it has to make 10 calls. This can take a few seconds.
To keep the traffic to twitter.com relatively low, we cache the result for up to one hour. So if you follow or unfollow people, it may take as much as an hour for the OPML to reflect the changes.
You could use this format to import your subscription list into another app, if one existed. As far as I know none does exist. However, this is an essential step in a bootstrap to create new ways of consuming Twitter data.
I'll have more notes here as they develop, and please post questions if you have them.
I sent a preview link to Anil Dash and here's what he said: "Heh. This got one of those immediate belly laughs that only come from seeing something new and realizing exactly how disruptive it can be. :-)"
Matt Mullenweg had the idea to import his Twitter subscription list into Google Reader. It worked, but all the subscriptions imported at the top level, meaning he had a clean-up to do.
So I added a feature to the app, if you add "&folder=1" to the end of the URL it creates an extra level in the OPML, designed for import into Google Reader (and probably other RSS aggregators as well). Example: