He seems like a brilliant fellow and is demonstrating some advanced threading concepts, only using distributed machines using DRb. This is an entirely different approach from a NeverBlock, which used Fibers, i.e. continuations, to allow green threads to "work" faster. I believe that based on the code I have seen today, Mike's approach is probably easier to understand and use.
He is also describing "Paxos" and encouraged us to take a look at Google's white paper "Paxos Made Live". Paxos is slow because there is lots of network chatter and Mike decided to head in a less formal direction using a designated "leader" and follower workers. Quote from his blog:
TokenWorkerensures that one process from a group of redundant processes will be elected to perform processing for the next N seconds.
StaticQueueWorkerdistributes a predefined set of work to a dynamic set of processes every N seconds.