No. 26 · Nov 12, 2026 · Response
Where do you look first during an incident?
Recent deploys are a good start. Just don't stop at your own.
It's 11 pm, and checkout requests are taking 8 seconds instead of 300 ms. You check what was recently deployed to checkout, and the last deploy went out two days ago.
So the checkout code hasn't changed, and the database looks healthy. You add a few pods and read through logs that show nothing new.
But the inventory team shipped a release 20 minutes earlier. Their new stock check runs a slower query, and checkout calls it on every order.
That's why the incident shows up on checkout while the change sits in another service. Your own deploy history can't show you that.
So it's worth looking at every deploy from the last hour, not just yours. A release in a service you call is often the context you're missing.
If you want to read more about how on-call responders work through an incident, it's all in this guide.