Torzon Status Bulletins

From roughly 09:15 to 09:55 UTC, Mirror C answered but at ~410 ms — above its 280–360 ms band — consistent with congestion on the CDN front the meek-azure path depends on. The endpoint stayed reachable throughout, so it was logged as slow, not down. It self-cleared without intervention. Users in that window were advised to prefer Mirror A or B.

Mirror B showed intermittent timeouts over snowflake for about 22 minutes from 02:48 UTC, traced to broker churn in the volunteer WebRTC proxy pool rather than anything at the endpoint. The control probe continued to reach Mirror B over obfs4 throughout, confirming the issue was transport-specific. Reachability returned to nominal once the proxy set re-stabilised; this event accounts for the dip in Mirror B's 30-day figure.

Probes from the West Asia vantage point began timing out on obfs4 to all mirrors, while the unfiltered control stayed green — a textbook regional block, not an outage. The recommended action for users there is meek-azure to Mirror C, which the matrix now reflects. No endpoint was affected; this is filtering at the network edge.

May closed with aggregate reachability of 99.7% across the three mirrors, counting a sweep as up when an endpoint answered from at least one region. Two short degradations and one regional block were logged; no full three-mirror outage occurred. Methodology is documented on the monitoring page.

A scheduled upstream maintenance window in Region A produced brief reconnection blips on Mirror A earlier in the week. As of 11:05 UTC on 28 May the endpoint was fully stable again, and it has recorded no outage since — the reason Mirror A leads the 30-day table at 99.95%.

A fourth censored-region probe came online in Eastern Europe, joining East Asia, Central Asia and West Asia alongside the Western Europe control. More vantage points mean the region matrix distinguishes regional blocks from genuine outages with greater confidence.