MidBehavioural

Walk me through a time you inherited a system nobody understood. Where did you start?

What they are really testing: How you build understanding under uncertainty, and whether your instinct is to rewrite. Proposing a rewrite in the first month is the answer that fails this question.

A real interview question

Walk me through a time you inherited a system nobody understood. Where did you start?

What most people say

drag me

It was such a mess that I proposed rewriting it properly from scratch.

A rewrite of a system nobody understands cannot preserve the behaviour that matters, because the important parts are exactly the undocumented edge cases the original accreted. Most such rewrites finish late and reintroduce old bugs.

The follow-ups they ask next

  • When is a rewrite genuinely the right call?

    When you understand the behaviour well enough to specify it, the platform is genuinely dead-ended, and you can run old and new in parallel comparing outputs. Understanding must come first either way.

  • How do you handle pressure to deliver features on it immediately?

    Negotiate a short stabilisation period with a concrete deliverable, framed in their terms: without observability the first feature change carries unknown risk to billing.

What the interviewer is listening for

  • Adds observability before changing anything
  • Learns from git and incident history
  • Explicitly resists the rewrite instinct

What sinks the answer

  • Proposes a rewrite early
  • Relies on documentation that does not exist
  • Changes many things at once

If you genuinely do not know

Say this instead of freezing. Reasoning out loud from what you do know beats silence every single time, and a good interviewer is listening for exactly that.

First I [made it observable, since silent failure was the immediate risk]. Then I learned from [logs, metrics, git history and past incidents] rather than docs. I [wrote the runbook as I went], made [small reversible changes like a dry-run mode], and [avoided a rewrite because the messy parts encoded real edge cases].

Keep going with behavioural

All 87 devops engineer questions

Knowing the answer is not the same as recalling it under pressure

Sign in to send the questions you fumble to spaced recall, so they come back right before you would forget them, and learn the concepts behind them with hands-on labs.

Start free