Random thoughts that spring up:
1) If DRK goes to, say, 10 bucks a coin in the mid-term, we are talking about 10.000 usd for running a node. It could be too much.
2) 10% of the mining for last node sounds too much. It also may open up a window of exploitation and give incentive to DDOS in master nodes so that one node takes the fee instead of the other... Other types of exploitation for getting the mining reward could spring up as we go along.
1000DRK is probably too much, it was just an example number. So was 10%. I was just vetting the idea itself.
Something else: The other day someone mentioned about the DRK allocation to various wallets and that Cryptsy has something like 400-500k DRK (?)... so, in this scenario, an exchange of that size can create 400 master nodes as a way to gain "interest" on their money despite the hot-wallet risk. Even if they only used 50% of their DRKs, it could be 200 nodes - which would be a sizeable part of the network under control of only 1 party. That would require quite a lot of laundering depth to avoid the chance of hitting a cryptsy node every single time.
That's a good point, the exchanges will definitely want to take part in those profits. I imagine they wouldn't want to run 100% of their coins as master nodes, that excessively dangerous. They must keep 80%+ in cold wallets.