Measurement methodology
Honest status numbers depend on an honest method. Here is exactly how the figures on this site are produced.
Live circuits, not cache
A prober opens fresh connections to each mirror over the real Tor network on a short interval. It does not read a cached result or ask the market. Each attempt records whether the onion answered and how long the round trip took. A single failure is expected noise on Tor and is not counted as an outage; a sustained run of failures is.
Latency is a distribution, not a number
"Reach" is reported as a representative round-trip time, but the underlying data is a distribution. Onion latency is naturally spread out because circuit quality varies. That is why the figure you see is a median rather than a single lucky or unlucky sample. When the network is congested, the whole distribution shifts and the median rises for every onion at once.
What the numbers do and do not tell you
| Signal | Meaning |
|---|---|
| Both mirrors reachable | The market backend is answering and at least two independent paths to it are healthy. |
| One mirror slow, one fast | A network path issue on the slow node. Use the fast one; both share one backend. |
| Both slow, rest of Tor slow | Network-wide congestion. Not a market problem. Wait it out. |
| Both unreachable for minutes, Tor otherwise fine | The likeliest real market outage. Check the incident log. |
Last reviewed 2026-06-28.