One of the differences that webmail services like Hotmail has is the ability that it does not deliver mail to the end-client, users have to login to their accounts and view their mail on the web (unless, of course, they POP their mail). Exchange Hosted Services, on the other hand, does deliver mail to the inbox and we don't store it where customers have to login to see (other than spam in their quarantine).
This ability to store mail for users gives Hotmail an advantage. If a spammer suddenly appears and starts hammering users with spam, many spam filters will fail to pick it up because the spam content is new and the sending IP has no reputation. It's hard to do reputation-based filtering on an IP that has no reputation, and it's difficult to filter content that you haven't seen before. What would be ideal is if after a reputation for the IP has been built up, if the IP is a spammer, to go back into the users' inboxes and if they haven't read the spam email yet, delete it. We may not catch the spam that has already been viewed by some, but we are catching the stuff from the same spammers that has not been viewed by others.
Recently, Hotmail has started doing this. I think they only started experimenting with it a few weeks ago so it's a very new technology (relative to Hotmail). I like the idea because if we don't catch a spammer at first, then at least we can go back and catch them later. We call this concept "Time-Travelling." We're effectively going back in time and fixing something that caused pain later on. Just like Hiro Nakamura.