He's running an individual exit node. You can run an exit node with as much bandwidth as you can spare. You can set bandwidth limits per month so that you don't go over your quota. Once your bandwidth's used up, people will just use a different one.
I'm running an exit node myself; I have a server that I only use for basic tasks that comes with a few terabytes of monthly transfer, so there's an IP on that server dedicated to Tor.
More seriously though, the price of bandwidth is dropping. I see these kinds of actions really as a way to prep the record with "stories" which link TOR and "Child Pornography" so that at some future proposal to outlaw such things a legislator will Google it and see a bunch of hits and go "wow this really is a problem, we should do something" having never read a single story.
If you're aware of this technique you can see examples of it all over the place. Another (perhaps local) version of this is MDMA which isn't technically illegal (its just a chemical that oh by the way if you ingest it affects you kind of like meth) and so first they label it "bath salts" then there are a lot of stories with "bath salts" + "bad outcome" (regardless if there is a proven or even plausible connection) which I expect to be followed by some sort of legislative action where all of a sudden I need to ask the pharmacist to buy Tums or something which can be used as a precursor to making it [1] from behind the counter and only in small amounts.
[1] The exemplar being sudafed which is controlled like crack cocaine in some jurisdictions.
Yes, people have published papers doing this:
http://homes.cs.washington.edu/~yoshi/papers/Tor/PETS2008_37...
My understanding is that looking at exit node traffic individually rather than statistically is likely to be illegal in many jurisdictions.
You could go one step further and play with MITM as well.