Match Release Strategy to Risk Profile

For each of the following three product scenarios, recommend a release strategy (shadow duration, canary sizing/targeting approach, ramp pace, and automatic-vs-human rollback posture), and justify each choice by explicitly naming the scenario's position on the volume and stakes axes:

1. A B2B SaaS company's new model recommends which of 200 enterprise accounts (all told) a sales rep should prioritize calling this week. A wrong recommendation wastes a rep's time on a low-value call; it does not lose a deal by itself.
2. A mobile game's matchmaking model pairs players in real time, 100M+ matches per day. A bad match mostly causes a slightly less fun game session; players who have consistently bad matches churn from the game over weeks.
3. A hospital's model flags patients for early sepsis intervention based on vitals, evaluated continuously for every inpatient, roughly 8,000 patients/day across the hospital network. A missed flag can be fatal; a false alarm causes unnecessary intervention and alarm fatigue among clinicians.