How would you answer an interview scenario involving Kubernetes probes?

A service takes two minutes to warm its cache and is repeatedly killed during startup. Explain how probe configuration should change.