I think that's the post you wanted to link to:
https://bitcointalk.org/index.php?topic=1958.msg612898#msg612898Only having skimmed their paper so far, but what they are getting at is that mining nodes might want to prohibit transaction with fees attached from spreading to other mining nodes. If a miner does not broadcast such a transaction to the network, he reduces the chance of other miners including these transactions into their blocks, thereby increasing his own expected reward over time.
Looking at how the network is actually structured today, this is a mere theoretical problem. The Bitcoin client goes to some lengths to ensure it is well connected to the network. As long as a miner does not somehow manage to encircle a client (controlling all nodes it is connected to), a miner does not really have a chance to prevent a transaction from spreading to the whole network.
Such an attack - however unlikely - should probably not be dismissed completely and I still have to read the whole paper to look at their proposed solution.
In my opinion, such a problem should be dealt with from the side of the sender of a transaction (who obviously has the greatest incentive for his transaction to be known by all miners). As soon as this issue would really start to become a problem, a client could simply spend more effort to make sure his transaction has spread.
