The business mandates a code freeze for all of December, the highest revenue month. Engineering objects. Who is right?
Show the full answer Hide the answer
What the interviewer is testing
Whether you can engage with a risk control on its own terms and offer something better, rather than either complying silently or dismissing it.
The business is right about the goal, wrong about the mechanism
The goal — protect the highest-revenue month — is sound and non-negotiable. The mechanism has a failure mode the business has not been shown.
Risk does not pause; it accumulates. Every change that would have shipped in December ships in January, as one large batch. That batch is more entangled, harder to review and harder to diagnose than any of its parts would have been. So the freeze converts many small independent risks into one large correlated one, positioned immediately after the window it protected — and January is not a risk-free month either.
Two secondary effects: the team loses the muscle memory of routine deployment right before the period where an emergency fix is most costly to get wrong; and urgent changes still happen, but through an exception process that has less scrutiny than the normal pipeline, which is the opposite of the intent.
What to propose instead
A tightened risk posture for the period, which protects the window without building the batch:
| Control | Normal | December |
|---|---|---|
| Batch size | Normal | Smaller, single-purpose changes |
| Canary bake time | 30 minutes | 4 hours |
| Rollout | Standard waves | Slowest wave order, one region at a time |
| Rollback | Available | Armed and automatic on any SLO breach |
| Change types permitted | All | No schema changes, no infrastructure changes, no dependency upgrades |
| Approval | Standard | Named engineering approver for anything touching checkout |
And a genuine narrow freeze where it is warranted: the specific systems in the blast radius of peak trading, for the specific peak days, not the whole estate for a month.
What a strong answer adds
Offering to measure it: run the tightened posture on the non-critical portion of the estate this year and compare incident rates against last year's freeze. That converts a belief into evidence and gives the business a way to change its mind without anyone losing an argument.
Also: whatever is agreed, the emergency path must be rehearsed during the freeze, not theoretical.
Common weak answers
"Freezes are an antipattern" with no alternative. Complying and then routing urgent changes through an exception path that bypasses the controls.