I feel like a complete jerk for saying this (partly because Sunlight Labs is a non-profit), but it suffers from what I like to call the snippet marketing syndrome: 4 - 5 lines of code being fancied up into a 10 - 30 line "plugin" with heavy references to the consulting studio that 'coded' it.
The idea behind Rack middleware is that it's incredibly easy to implement for the developer and re-usable across the spectrum of Ruby web apps.
Yes, implementing Google Chrome Frame in a given app is trivial. But removing this from the concern of a given app and into a re-usable middleware is where I think it's useful.
I think the biggest disadvantage of putting something like this in middleware, is that it's completely incompatible with stuff like memcache. The injection of this meta-tag is done after a memcache SET would have been done in the web app, and so if you have a reverse proxy in front of your webserver, for any cached page that's served, it won't include the injection. Also, the user-agent check that's being done in this plugin yields incorrect results in the presence of caching for similar reasons.
