Joining a New Team Overwhelmed by On-Call Fires: What Do You Do?

This question closed a domain deep-dive round. The candidate first presented a few slides on their past work and answered questions about it; the interviewer then moved from past experience to a hypothetical: > You are placed on a new team whose on-call rotation is in constant firefighting: many pages, recurring incidents, and little time left for planned work. What do you do? Answer the hypothetical directly, as a plan, and support it with evidence from your own experience where you have it. ```hint Diagnose before you fix Before proposing changes, decide what information would tell you where the fires actually come from and which of them cost the team the most. ``` ```hint Two time horizons Separate what keeps the service and the on-call engineers afloat this week from what removes the causes of the fires over the next months, and say whose agreement the second needs. ``` ### Clarifying Questions - Do I join as an individual engineer or as the team's lead, and am I on the rotation from the first week? - What counts as a fire here: pages, customer-facing incidents, escalations from other teams, or support tickets? - Is the load new (after a launch, a growth spurt or a dependency change) or long-standing? - Does the team already keep incident records, runbooks and postmortems, or do those need to start from scratch? ### What a Strong Answer Covers - Clarifies its own role and the nature of the fires before prescribing anything - A first phase of learning: joining the rotation, reading past incidents and measuring the load by source - Immediate relief that does not hide real problems, such as alert cleanup, runbooks and faster mitigation - A prioritized, data-backed plan to remove the main recurring causes, with time for it agreed with the manager and partners - Attention to the team's sustainability and to spreading knowledge across the rotation - A way to measure and report progress, and a real past example that shows the same instincts ### Follow-up Questions - Your manager says the roadmap cannot slip this quarter. How do you get time for reliability work? - The noisiest service belongs to a senior engineer who believes every alert is necessary. How do you handle it? - Three months in, pages have halved, but a new kind of incident appears. What do you change? - How would you know your plan worked, and what would you report upward?