bridge authority could do bandwidth test along with reachability test
It's increasingly clear we want to know which bridges are slow, so we can treat them differently (e.g. never give out the slow ones).
I heard Mike and Aaron talking about implementing a remote bwauth style thing to remotely measure how much throughput they could get from each bridge.
But I think we don't need an accurate bandwidth estimation -- we need to know if they're too slow or not too slow.
So it occurred to me that Tonga's reachability test is nearly doing the test we need. If it pushed some traffic down its reachability circuit, maybe in a way that triggered some traffic in response or maybe just by making a loop and sending the traffic back to itself, then it could see what it gets, and put that number as the bandwidth weight in the bridge networkstatus document, and bridgedb could just use it.
I'm normally not a fan of complexifying Tor when some outside service could be written to do the same thing. But we're so close! :)
Another argument in favor is that to limit blocking risks, we want to limit the number of points on the network that probe the whole set of bridges.