fail-transport-class targets one authored per-node connectivity profile such
as tor, bluetooth, or wifi. At its virtual timestamp, every edge incident
to an assigned node in that class becomes unavailable in stable edge-ID order.
restore-transport-class removes that failure block later.
Availability blocks compose. An edge affected by both a network partition and a transport-class failure does not reactivate until both causes are restored. Each transition records its assigned nodes and the edges whose visible state actually changed.
Failed edges leave the normal routing graph. Parent/root repair, Bloom peer state, coordinate-cache invalidation, session disruption, and later synthetic payload routing all use that updated graph. Restoration requests fresh tree and Bloom exchange on every reactivated edge.
The class assignment and failure semantics are exact configured inputs. Bandwidth, latency, jitter, loss, MTU, and queue values remain modeled rather than measurements of a real transport implementation.
The native workbench exposes profile selection plus fail and restore buttons at the timeline cursor. Changed edges dim immediately during replay and the inspector counts currently failed classes.