Interesting! I'm thinking the clusters are at points that the plates bind, based on the fact that the fault plane is an irregular surface. I don't think it's fluid migration. Still pondering why they are going deeper over time ... maybe the binding points wear down over time since the previous large quake, hence the load transfers downwards as there is less and less friction from the binding points in the upper part of the fault .... Is the rock in the upper part of the fault more liking to create binding points after a large quake rather than the lower portion of the fault?

Michael