Phase 67 · multi-root / load-balanced address fabric · AT-0641 → AT-0656
One root chokes. Many roots breathe.
Phase 66 left multi-user scalability unresolved and found a root-band deadlock in the single-tree fabric. The structural answer is to stop electing one root. The substrate builds K generic BFS trees before any endpoint exists, every node stores K compact O(log N) addresses, and each trip picks a tree using only the addresses it already holds plus a congestion counter at its own node and first hop — no route oracle anywhere. This page runs the exact event-driven simulation on the real cubic graph, and keeps the closed-form queueing estimate quarantined in its own panel where it can predict but never report.
Plain language
Loading.
Headline readings · exact layer
Completion gain, K=1 → K=4
…
F = 16, N = 256 … 4,096
Busiest-node load
…
sublinear in offered load
Active footprint
…
on top of O(log N) route
Multi-user scalability
OPEN
Running the exact multi-root simulation in your browser…
Verdict
…
Status of the programme after this phase is unchanged: LOCALIZED AUTONOMOUS DOORWAY — TOY CANDIDATE / PHYSICAL EVIDENCE NONE. An architecture that disperses traffic is not an architecture that guarantees delivery, and the gate stays open until it does.
Results · AT-0641 → AT-0656