Early on I thought the best engineer in an incident was the fastest one. Most tabs open. Hotfix shipped. Name in the retro.
Sometimes that person helps. Often they are just loud. Loud is not the same as right.
What actually ends incidents is quieter. Someone who slows the room enough to say what we know, what we do not, and what we will not try yet. Someone who keeps the blast radius small. Someone who writes the timeline down while adrenaline is still inventing a cleaner story.
What calm is not
Calm is not not caring. Users are still hurt. On-call is still tired.
Calm is not waiting forever. You still ship the stop-the-bleed change when the evidence is good enough.
Calm is refusing to turn fear into a second outage. Synchronized retries. Bad deploys. Blame that teaches the next person to hide what they saw.
What I practice now
After years of systems that fail under real load, this is what holds:
- Stabilize first. Skip the known bad path. Shed load. Then improve the design.
- Say the uncertainty out loud. “We think X. We have not ruled out Y.” People relax when the map is honest.
- Protect the next human. Leave a note the 3 a.m. version of your teammate can use. Mentorship is not only 1:1s. It is incident hygiene.
What I still believe
Seniority shows up as anticipated consequences, including social ones. A technically valid change that panics the team or erases the evidence is not responsible engineering. The system includes the people on call. If your heroics teach them that speed beats structure, you trained the wrong lesson for the next page.