Preventing a Cache Stampede After a Deploy

After every deploy, your Redis cluster restarts cold (no persistence configured). Within seconds, the database sees a 40x spike in query volume and its CPU saturates until the cache warms back up naturally, sometimes causing a brief outage.

1. Explain precisely why a cold cache produces a spike this large rather than a gradual ramp.
2. Propose two complementary mitigations, one that changes what happens before traffic arrives and one that changes what happens during the miss storm, with the mechanism for each.
3. Would increasing the database's read replica count fix the root cause? Why or why not?