Imagine there are no miners. This is the practical endgame result when the chance of solving a block drops so low there is no point for ordinary users to be a miner.
Why would I run a Bitcoin client then? What are my incentives?
The health of the bit coin network or maybe society? There will always be some that do this. But most people value short gain profit over anything else.. Ref 'tragedy of the commons' or for that matter 'environment problem X'.
Perhaps I will run it because I use Bitcoin myself? I would certainly use it in the moments I need access to Bitcoins.
But in the 'idle' periods? I have no incentive, this would just be another bandwidth user and potential entrance into my system.
Give me incentive. Give me a chance of 'earning' something running my client and I would probably run it. But in my scenario, remember that there is no practical chance of getting a payment for a new block, so I would be interested in earning the transaction fee instead.
But once you have incentive for doing something, you run into the Sybil problem the Microsoft researchers try to defeat with their hybrid model. Assume zombie networks with lots of machines running for this purpose, and they might be able to at least push their odds in their favor.
The Microsoft analysis seems to be a sound analysis, given some assumptions. I believe that we, as technical people, should do less posturing and more scientific method. If you disagree, see if you can disprove that their assumptions were correct or try to prove a different model. I happen to agree.