Is that part of the robots.txt RFC or just a happy security coincidence? The bit.ly issue seems like a pretty bad bug that ought to be fixed.
You also might be able to fool Google Webmaster Tools or other utilities into thinking you own bit.ly. One of the authentication methods I've seen used was the creation of an empty html document in the root directory.
The 1994 original proposal is silent on what to do with any unexpected response codes/types.
The 1997 internet-draft (never an RFC) suggests the redirect should be followed, but the potential confusion that could cause has meant few, if any, crawlers have followed that guidance. It's more likely to be a webmaster configuration error than real intent.
"And what about sitemap.xml, atom.xml, and other typical files that could also be redirected?"
Why not try it? As of this writing, both of those go to other places.
I checked the other one that lept to mind, favicon.ico. Bit.ly appears to have hardcoded it (probably in apache configuration or equivalent for whatever server they use), however, try http://bit.ly/faviconico and look at the resulting URL. Looks like a few things were tried by the same guy. Now, pwning the favicon would have been cool.
You forgot: Lotus Notes. Mishandles html emails so much, the new Outlook is pure perfection compared to it. (grumbles over having to implement a complex receipt email for a company that uses Notes internally).