I have a few questions related to how bandwidth scanners work. I would appreciate it a lot if anyone could answer any one (or all) of them:
From what I read in the bandwidth scanners spec, my understanding is that nodes are first divided into subsets of fastest, fast, slow and slowest. Then each of the bandwidth scanners probe these subsets individually by creating 50-node slices, and creating a circuit using two nodes from each slice. I am confused about how relays are divided into 50-relay subsets. Are they spread out in percentiles according to their bandwidth just as the categorization of fastest, fast, slow and slowest criteria. And are these 50 relays chosen from the subset already assigned to the particular bandwidth scanner or they move across bandwidth scanners.
Also, I was wondering how directory authorities isolate the bandwidth of a single node using a two-hop circuit. Shouldn't the bandwidth measured using the two-hop circuit give an approximation for the entire circuit rather than the node?
From the specs it seems that the bwauths have specific criteria for choosing nodes for these two-hop circuits (having Fast and Running flags, leaving out subsets with no exit nodes, /16 subnet different requirement). Is there ever a possibility that relays are left out? And some relays not measured at all.