What worked
The tactile controls, restrained visual system, deterministic relationships, explicit limits, and browser-local progress created a coherent research artifact.
Archive / replaced experiment
A polished interface can still operationalize the wrong idea.
The clinic outage experiment is preserved as design evidence, not as a live legacy application. It helped expose the difference between a reusable reasoning engine and a reusable interface.
Why it was replaced
A system-operation task stood in for bidirectional comprehension. The permanent control surface, route checkpoints, scene taxonomy, and five-role trace made the mechanism more visible than the learning idea.
What carries forward
Stable authored relationships, retryable feedback, an inspectable proof, accessible DOM controls, and the exact tactile sound remain part of the lab.
Architectural conclusion
The evaluator stays shared. The visible instrument does not.
V2 and V3 remain recoverable in Git history. No clinic components, progress, feedback, or simulation are loaded by the new experiment route.