Fetch networkstatuses scalably
Right now we try to fetch the minimum number of networkstatus pages that would make us happy. If one fails, we wait a minute before trying just enough again. We should do something smarter, for example, fetching more originally, and retrying immediately to fetch another pile if too many fail from the first attempt.
Along with this, we need to audit all the places in the code where we require a threshold of networkstatuses to claim something. I believe we have fencepost errors all over. And in some places we still use constants that assume there are always 3 networkstatuses in the network. Lastly, we should make sure that an independent Tor network can still function if it has only one authdirserver.
[Automatically added by flyspray2trac: Operating System: All]