A bug in Tor's nodelist handling caused clients to permanently mark an exit relay as rejecting all traffic after receiving a single transient RELAY_END message, such as RESOURCELIMIT or HIBERNATING. This issue, introduced in Tor 0.2.3.1-alpha, allowed a third party to steer clients away from honest exits by exhausting an exit's sockets or sending a malicious END. The fix, which clears the 'rejects_all' flag when a new consensus is installed, is slated for the next stable release.
Why it matters This fix prevents a remote attacker from manipulating a client's usable exit set, which could otherwise lead to targeted denial of service or privacy leaks by biasing circuit selection.
Across 2 notifications dated 2026-11-09T14:07:20Z through 2026-11-09T15:07:27Z, Tor Project Consensus Health Monitoring Mailing List reported a missing authority descriptor for moria1. These are reports from a monitoring feed; the notifications do not establish a continuous outage, its cause, or its operational impact.
Why it matters The repeated document checks provide a source-linked monitoring observation for further investigation; the cause and wider impact remain unestablished.
Across 3 notifications dated 2026-11-08T22:07:20Z through 2026-11-09T02:07:20Z, Tor Project Consensus Health Monitoring Mailing List reported a missing authority signature; an outdated consensus document for dizum. These are reports from a monitoring feed; the notifications do not establish a continuous outage, its cause, or its operational impact.
Why it matters The repeated document checks provide a source-linked monitoring observation for further investigation; the cause and wider impact remain unestablished.
Across 16 notifications dated 2026-11-08T10:09:18Z through 2026-11-10T04:07:22Z, Tor Project Consensus Health Monitoring Mailing List reported a missing authority in a consensus; a Guard-flag discrepancy for maatuska. These are reports from a monitoring feed; the notifications do not establish a continuous outage, its cause, or its operational impact.
Why it matters The repeated document checks provide a source-linked monitoring observation for further investigation; the cause and wider impact remain unestablished.
Across 20 notifications dated 2026-11-08T09:13:49Z through 2026-11-10T07:13:46Z, Tor Project Consensus Health Monitoring Mailing List reported a missing authority in a consensus; a missing authority descriptor; a missing authority signature for bastet. These are reports from a monitoring feed; the notifications do not establish a continuous outage, its cause, or its operational impact.
Why it matters The repeated document checks provide a source-linked monitoring observation for further investigation; the cause and wider impact remain unestablished.
Correction · Corrected an unsupported claim that ordinary relay operators must edit their configurations. The cited issues request removal of bastet's existing entry from the software.
The Tor development team reported that the operator of the bastet directory authority notified them that the machine is very plausibly compromised. The report states that bastet will be removed from the source code because its keys can no longer be trusted. The operator is expected to generate a new identity for the authority.
Why it matters The report says bastet's existing keys can no longer be trusted and requests removal of its current entry from the software.
Following a period of packet drops and consensus delays affecting the maatuska directory authority, Tor 0.4.9.16 was released. The release introduced consensus-method 38, which reportedly eased pressure on directory authorities and reduced traffic from maatuska. Upstream network issues that contributed to the overload were also addressed during this period.
Why it matters The release of consensus-method 38 is intended to ease pressure on directory authorities.
Tor developers merged a revert of a recent change that listed multiple protocol capabilities as required. The revert addresses issues causing clients on version 0.4.8.x or lower to get stuck in a restart loop while fetching unusable consensus. This action allows affected clients to rejoin the network and treat microdescriptions as valid.
Why it matters The revert is intended to resolve a network-wide incident where a significant number of clients were unable to connect due to consensus fetching failures.
A community report on the tor-relays mailing list presents custom metrics from approximately 300 relays, identifying 30 distinct flood windows over 48 hours. The data indicates that descriptor fetching and HTTP-wait closures dominated the traffic, with one relay recording over 14 million fetches. The report notes that Tor 0.4.9.15 is expected to rate limit some of these requests, while a separate incident report describes a relay experiencing memory exhaustion and OOM termination during a similar period.
Why it matters The detailed metrics provide specific evidence of the traffic composition during recent network-wide resource exhaustion events, helping to distinguish between different types of flood attacks and validate the effectiveness of upcoming rate-limiting measures.
A relay operator reports a sustained circuit-building DoS against their guard relay family starting June 25, 2026, characterized by an 8x sustained increase in open circuits and a 20x peak. The attack involved millions of near-idle circuits and onionskin processing spikes across multiple regions, with bandwidth increasing only 20%. While DoS defenses engaged and prevented TCP port exhaustion, EU relays shed 10-20% of real traffic due to capacity constraints.
Why it matters The report suggests the pattern is network-wide or family-targeted, and notes that EU relays shed 10-20% of real traffic.
Following a power outage on 2025-10-27, the Tor directory authority maatuska has been reporting substantially fewer relays with Guard flags in its votes compared to the network consensus. While initial reports showed a discrepancy of over 3,000 flags, recent data from November 2026 indicates the gap persists, with maatuska voting for roughly half the Guard relays of other authorities. The issue remains open.
Why it matters Scope and consequences remain subject to the linked upstream discussion.
