Regarding the probability of attack, people should monitor the number of different usernames attempted by a session/IP not just failed attempts against individual accounts. Otherwise it is very easy to try thousands of username combinations with a selected weak password.
"But after some further consideration, we decided that it was a false risk, as the username reminder form already tells you if a username exists"
Emails are more personal and might be easier to link back to personal information. Thus, confirming that there is an associated account with a given email is also a privacy leak, because maybe people don't want to reveal that they have an account on a specific website.
And as they say in the article, that information is already leaked by the password reset process: "No email by that name available".
This is an old pattern. The reason for it seems to have been forgotten: If you tell the user they got the user name wrong, you're leaking information. They can now try and guess valid account names. Of course, one could argue that this may not matter in their specific case, but it does in some.
The parable of chesterton's fence comes to mind: http://en.wikipedia.org/wiki/Wikipedia:Chesterton%27s_fence
edit: Ah, I really need to overcome that instinct of jumping to the comments whenever I notice something. Looks like they addressed this.