Treat a long incident as a relay, not a sprint. Split the roles, hand command over at the end of each working day with an explicit acknowledgment, keep one living document, and make sure the person who has been up longest goes home. Tired responders make slower and riskier decisions.
Why tiredness is a security risk
Google's on-call chapter says stress hormones can impair cognitive functions and cause suboptimal decision making, and that people under stress may fall back on quick habitual reactions that are risky when unexamined. In an incident, that means skipped steps and missed evidence.
Split the work
The SRE incident chapter describes four roles.
- Command holds the overall state and structures the team.
- Operations is the only group changing the system.
- Communication is the public face and sends updates.
- Planning files bugs, arranges handoffs and tracks drift from normal.
When one person holds everything, they tire first. The chapter says that a member whose load becomes excessive should ask the planning lead for more staff and delegate, possibly by creating subincidents.
Plan the relay
- Decide at the start roughly how long the incident might last.
- Name the next commander early.
- Hand off command at the end of the working day.
- Brief the incoming commander first, by call if remote.
- Wait for a firm yes before the outgoing person leaves.
- Announce to all responders who is now in charge.
The chapter's own example has the team finding replacement staff before going home at 5 p.m. and briefing colleagues in another office.
Protect rest
- Set a maximum continuous time for any responder. Pick a number with your team, since the source gives none.
- Make leaving an explicit decision by the commander, not a favour.
- Provide food and quiet space if people are on site.
- Excuse people from routine work the next day.
Keep the record
The commander's most important job, the chapter says, is the living incident document. Put the key facts at the top. It lets a fresh person take over in minutes and feeds the postmortem.
After the incident
Plan the recovery as carefully as the response. Give responders time off, and make sure the review happens after people have rested, not the next morning. Reviews held by exhausted people tend to blame.
Say thank you in a specific way: name what each person did that helped. Recognition after a hard incident matters for morale, and the postmortem chapter recommends publicly recognising people who handle incidents well. It also builds the habit of declaring incidents early next time.
Quick check
Should the incident commander also fix the problem?
Avoid it. The commander holds the picture and removes roadblocks. If the commander dives into fixing, nobody is watching the whole incident.
Sources
Last reviewed: 19 October 2026.
