The concept is this: Master nodes, are in effect the decentralized "laundry" nodes where the mixing takes place. It has been suggested that the cost to run one should be high enough that large entities that want to map the network do not saturate it, so a requirement of ~1000 DRK will probably be programmed as a requirement to run a node.

Now, as a form of incentive for running the node (and compensation for having associated costs + potential risk to one's 1k DRK) it has also been suggested that aside from transaction fees these nodes could be getting like 10% of the mining output per block, given to the last node which processed a DarkSend. This would allow honest people to run nodes due to the incentive, instead of allowing the network to be saturated by controlled nodes from an entity hell bent on mapping the entire network. The DarkSends will probably be processed by multiple nodes, so even if one is controlled by an entity wanting to monitor transactions and the next node is not controlled = the monitoring breaks.

Now in terms of economy: If, say, 3m coins are produced a year, 300k will go to DarkSend Nodes. If there are 300 nodes with 1k invested, it could make these 300k => 600k within a year, with no mining, but simply contributing a node. It's an interesting new way to make money by having the incentive to provide a mixing service while also gaining "interest" for investors/bagholders etc. This could be large in the future. It's actually proof of work (work=mixing) but also requires a stake to participate.