A cache miss on a popular object is not one miss. It is every request that arrives while the first one is still fetching, and on a cold start that can be the entire front page of a site at the same instant.
The lease
Avand's edge cache takes a short-lived lease on the key before it goes to the origin. The request holding the lease fills the cache; the others wait on it and are served the result. Nobody else calls the origin, so an origin sized for steady traffic is not asked to absorb a stampede it was never provisioned for.
Purges across replicas
A purge that only clears the replica that received it is worse than no purge, because it looks like it worked. Purges are published over Redis pub/sub and every replica drops the key, so the result is the same whichever node the request lands on.
What this means in the panel
The purge confirmation says what it will affect, and a full-zone purge asks again. That is not caution for its own sake: refilling a large zone from origin is a real cost, and it should be a decision rather than a reflex.