How do I stabilize online services after a Redis memory alert?

I was once on duty for an e-commerce backend during a flash sale event. The problem didn't seem major at first: Redis memory usage suddenly spiked to 92%. The API hadn't crashed yet, but cache misses and response times were already fluctuating. Everyone on-site had a reasonable explanation, but when put together, they didn't add up. I first asked everyone to stop taking actions that might…