NM-003: Circuit behaviour and reachability failures

1. Failure classes

Reachability failures fall into three classes and they need different responses. Confusing them wastes time at best and produces a phishing loss at worst, because the reaction to a perceived outage is where most losses begin.

2. Circuit renewal

A circuit is the path your traffic takes through the network and it degrades over time or lands on a slow relay by chance. Requesting a new circuit for the site rebuilds that path and resolves the majority of local failures. If it does not, closing the browser entirely and reopening rebuilds everything rather than one path.

3. Service side pressure

Sustained floods against hidden services are routine. During one, an address may accept a connection and then stall, or refuse entirely. This resolves on its own timescale, usually minutes, and there is no user side action that speeds it up.

4. Prohibited responses

• Do not search for a replacement address during a failure, which is the sequence behind nearly every credential loss around markets
• Do not lower the browser security level, which has no effect on reachability
• Do not install alternative clients or accelerators claiming to improve onion access
• Do not conclude that a market has ended because one address was unreachable for an afternoon

5. Escalation threshold

A single address quiet for minutes or hours is unremarkable. Every verified address unreachable for days, with no signed communication anywhere, is a different situation and justifies caution rather than action. Even then the correct response is patience, because the alternatives available during an outage are overwhelmingly hostile.