Reach-Weighted Review Prioritisation

Your cascade's automated layers leave 800,000 assets/day in the ambiguous middle. Human review capacity is 25,000 assets/day. Two assets both have a calibrated synthetic-probability score of 0.55: Asset A is a video from an account with 40 followers; Asset B is a video from an account with 2 million followers whose view count is climbing rapidly.

1. Explain why a policy that routes strictly by raw score, highest first, is not the right design here, using these two assets.
2. Propose a concrete prioritisation formula and show how it reorders A and B.
3. What operational risk does reach-weighting introduce that a pure score-based ranking doesn't have, and how would you mitigate it?