OK, I will make tests with Floyd Warshall algorithm, and also with Dijkstra’s algorithm.
The tests will consist in a random distribution of 16 links per member, with 1.000.000 members. And test roughly 2% (this is worldwide birth rate) random members for distance.
I will be able then to make the links vary in quantity and also the quantity of members.
The problem is not the frequency of the block, but the frequency of the newcomers/renewers. Each generates a distance test against the WoT.
Also, the cost is in the extraction of the links to create a WoT matrix that can be computed in-memory.
Finally, the current rule still have a big problem as an attack has been described here. So the current distance rule cannot be applied to the whole WoT at any time, but a part of it.