For users who suspect a specific mirror is down, the per-mirror state right now plus the diagnostic flow if your specific Tor circuit can't reach a mirror that the probe says is up.
The live per mirror state is on the status board, and it changes. What stays true is the shape of the problem below. If your Tor circuit cannot reach a mirror the probe says is operational, the failure is local to your circuit topology rather than the mirror itself; the recovery flow is in the “not working” landing page on this site.
Mirror-down false positives are common during Tor guard rotation events. The pattern: a guard rotation on the Tor side moves your client to a new entry guard whose path through Tor to the Nexus production guards is congested or fails outright; you see the mirror as down, the probe (running through a different guard) sees it as up. Wait five minutes, refresh circuits, retry.
This is precisely the scenario the address set exists for. Move to the next one and carry on with the same account and the same balance. There is nothing to investigate and nothing lost.
Tor moves your client to a new entry guard periodically. If the new guard's path to the service is congested, you see a mirror as down while a probe running through a different guard sees it as fine. Wait a few minutes, request a fresh circuit, retry. This accounts for a large share of single mirror reports.
Two fresh circuits failing the same way on the same address while another address answers immediately. That is a reasonable conclusion and the correct response is still to use the one that works.
There is no advantage to using a particular address. They are the same service and the only difference is the path. Spending twenty minutes getting a stubborn mirror to answer buys you nothing that the next one in the list would not have given you immediately.
Component states, the 90 day uptime history and the incident log live on the status dashboard. This page covers one question in depth rather than repeating the board.