A school-safe marketplace simulation tests a new revenue page. Halfway through, a popular teacher mentions the simulation in class, causing a traffic spike to one version only. Text exhibit: Experiment timeline showing Days 1-3 balanced routing to Version A and Version B, Day 4 teacher mention sends spike to Version A only, Days 5-7 marked review imbalance. Decision box says extend or rerun with balanced routing before applying preset rule. What should the team do to protect the experiment?