Technical leadership

Question 1

You lead the team that owns the payments ledger service, which twelve other teams call. A partner team asks you to add a field to the core ledger response that only they need. What do you do first?

Accepted Answer

Ask what problem the field solves and whether the ledger is the right home for it.

Question 2

Your identity service is called by thirty services. Callers keep filing tickets asking how retries, timeouts, and error codes behave. What is the most durable fix?

Accepted Answer

Publish a written contract for timeouts, retries, error codes, and limits.

Question 3

A team calling your inventory service sends ten times its normal traffic during a sale and degrades the service for everyone. What should you, as owner, put in place?

Accepted Answer

Per-caller quotas agreed in advance, plus a clear way for teams to request more capacity.

Question 4

You become tech lead of a notification service with no stated reliability target. Callers say it feels flaky, but nobody can say how flaky. What do you do first?

Accepted Answer

Measure current success and latency, then agree on a target with the main callers.

Question 5

Your search service team is small, and other teams keep asking you to build features in your codebase. You cannot keep up. Which approach scales best?

Accepted Answer

Accept outside changes, with clear guidelines, named reviewers, and contract tests.

Question 6

A caller of your pricing service relies on the order of items in a response, which was never documented. You want to change the internal sort. What is the right move?

Accepted Answer

Treat the order as contract for now, tell that caller, and agree on a date to change it.

Question 7

Two teams want opposite behavior from your geocoding API when they hit its rate limit. One wants fast rejections, and the other wants requests queued. How do you decide?

Accepted Answer

Pick a default that fits most callers, and offer the other as an option if it is cheap.

Question 8

Your team owns the order service. A failing downstream service caused your error rate to spike, and your callers blamed you. What do you change as owner?

Accepted Answer

Add timeouts and fallbacks for that dependency and show its health on your dashboard.

Question 9

You need to retire version 1 of an internal accounts API that about forty services still call. What do you do first?

Accepted Answer

Use traffic logs to find exactly which callers still use version 1, and how much.

Question 10

Three months into deprecating an old API, 35 of 40 callers have moved. The last five keep saying they are too busy. What works best?

Accepted Answer

Meet each of the five, offer real help such as a ready pull request, and agree on dates.

Question 11

A product manager wants a breaking change to a public order API shipped next week. External customers integrate with it. What do you propose?

Accepted Answer

Ship it as a new version next to the old one, with a published end date for the old version.

Question 12

Your team has argued for three weeks about whether to version an internal API by URL path or by request header. What do you do?

Accepted Answer

Pick one, write a short reason, and move on, since either works and consistency matters.

Question 13

Before you remove a deprecated endpoint, which step most reduces the risk of a surprise outage?

Accepted Answer

Run announced brownouts where the endpoint fails briefly, so hidden callers show up.

Question 14

An engineer on your team proposes making a new field required in a request schema that many services send. What is the best feedback?

Accepted Answer

Add it as optional with a safe default first, and require it later in a new version.

Question 15

You are writing the deprecation notice for an old reporting API. What must it include?

Accepted Answer

What is changing, the replacement, the dates, how to check if you are affected, and who to ask.

Question 16

Halfway through an API deprecation, a large caller says the new version lacks a feature they depend on. What do you do?

Accepted Answer

Confirm the gap, then add it or offer a workaround, and move that caller's date if needed.

Question 17

Your team keeps shipping small breaking changes to a shared events schema by accident. What lasting fix do you lead?

Accepted Answer

Add a check to continuous integration that fails any change that would break readers.

Question 18

You are leading a move of user profile data out of a shared database that six services write to directly. What is the first milestone?

Accepted Answer

Put one owning service in front of the data so the others stop writing to it directly.

Question 19

Your plan moves orders data from one database to another, and four services read and write it. Which plan shape is safest?

Accepted Answer

Dual write, backfill, compare reads, move reads one service at a time, then stop old writes.

Question 20

Two months into a cross-service migration, a comparison job shows 0.3 percent of records differ between the old and new stores. The deadline is in two weeks. What do you do?

Accepted Answer

Find the cause before moving reads, and tell stakeholders now if that pushes the date.

Question 21

A team whose service is part of your migration says they cannot do their share this quarter. What is the best response?

Accepted Answer

Offer to have your team write the change for them to review, and agree on a date together.

Question 22

You are writing the plan for a migration that touches seven teams. What makes it easiest for those teams to commit?

Accepted Answer

Each team's tasks, effort, and dates, plus what happens if they slip, agreed with leads.

Question 23

Halfway through migrating billing data, you find the rollback path was never tested. Reads have already moved for two services. What do you do?

Accepted Answer

Pause further cutovers and test the rollback for the two moved services first.

Question 24

After a migration, when is it time to delete the old database?

Accepted Answer

When nothing has used it for an agreed period, backups are archived, and owners sign off.

Question 25

During your cross-service data migration, an engineer suggests redesigning the schema too, since you are touching it anyway. How do you respond?

Accepted Answer

Keep the migration a like-for-like move, and plan the redesign as a later step.

Question 26

Your migration is on track, but leadership keeps hearing rumors that it is stuck. What prevents this?

Accepted Answer

A short weekly written status with progress, risks, and asks, sent to the same people.

Question 27

You are asked to set on-call standards for twenty backend services. Several teams page on every warning. What standard do you propose first?

Accepted Answer

Page only for problems that hurt users and need action now; ticket everything else.

Question 28

After a reorg, a payment retry service has no clear owner. It pages a rotating set of people, and nobody fixes root causes. What do you do?

Accepted Answer

Propose one owning team to engineering leadership, with reasons and a handover plan.

Question 29

Which on-call rule best prevents burnout while keeping services reliable?

Accepted Answer

Track pages per shift, and make a team fix its service when pages pass a set budget.

Question 30

You want every backend service to meet a basic production-readiness bar before it takes traffic. How do you roll this out?

Accepted Answer

Write a short checklist with the teams, apply it to new services, and give old ones a dated path.

Question 31

A team asks for an exception to the new on-call standard because their reporting service is low priority. What do you do?

Accepted Answer

Check that its callers agree, and if so, record a lighter standard for it with a review date.

Question 32

The on-call runbook for your core service is long and out of date, and new engineers struggle with it at 3 a.m. What fix do you lead?

Accepted Answer

Update the runbook in every incident follow-up, and test it with new engineers.

Question 33

Two teams argue over who should be paged when a shared message queue backs up. Each says the other owns it. What should you do?

Accepted Answer

Assign it to the team that can fix the usual causes, write it down, and check in a month.

Question 34

Your team's service rarely pages, so engineers forget how to handle incidents. What helps most?

Accepted Answer

Run regular drills on a staging copy, with rotating roles and a short review after each.

Question 35

A product manager wants a new payout endpoint live in three weeks. The fastest plan skips idempotency keys, so a client retry could pay someone twice. What do you do?

Accepted Answer

Explain the double-payment risk in money terms, and offer to keep the keys but cut other scope.

Question 36

After you explain the double-charge risk, the product manager still insists on the launch date. Your own manager does not know about the disagreement. What next?

Accepted Answer

Bring your manager a written summary of the risk and options, so the right level decides.

Question 37

Which way of stating the idempotency risk is most likely to change a product decision?

Accepted Answer

At our retry rate, about 40 orders a day would be charged twice and need refunds.

Question 38

Leadership agrees to keep idempotency in the payout launch and asks what else can be cut to hold the date. Which cut is most reasonable?

Accepted Answer

Launch with one payout method instead of three, and add the other two the following month.

Question 39

Leadership accepts the risk in writing, and the payout endpoint ships without idempotency. What is your role now?

Accepted Answer

Commit to the launch, watch for duplicate payouts, and make the fix the next item.

Question 40

A junior engineer says the payout deadline is impossible and asks you to just tell product no. How do you respond?

Accepted Answer

Ask them to help size the work and options, so you can bring product a tradeoff.

Question 41

Two weeks before launching an order import API, you find that client retries can create duplicate orders. The date is already committed. What do you do first?

Accepted Answer

Tell the product owner today, with the impact and options, before deciding if the date moves.

Question 42

Product keeps setting launch dates before engineering has scoped the work, and correctness corners get cut each time. What do you change?

Accepted Answer

Agree with product that dates follow a short scoping pass, with a range and risks.

Question 43

Finance flags that your backend platform's cloud bill is 60 percent over budget this quarter. What do you do first?

Accepted Answer

Break the bill down by service and resource to find the few items driving it.

Question 44

The cost breakdown shows that keeping full debug logs for a year costs more than all of your compute. Teams resist any change. What do you propose?

Accepted Answer

Keep debug logs a few days and error and audit logs longer, and show teams the savings.

Question 45

Costs on your account data service grew because one caller polls it every second for changes. What is your best move as owner?

Accepted Answer

Learn what the caller needs, offer push updates or caching, and set a rate limit going forward.

Question 46

Your team wants to buy reserved capacity to cut costs by 30 percent, but traffic forecasts are uncertain. What is a sound decision?

Accepted Answer

Reserve only the baseline you are confident in, and keep the rest on demand.

Question 47

Cloud cost overruns keep surprising your org months after the fact. What lasting change do you lead?

Accepted Answer

Tag resources by team, and give each team a weekly cost view with growth alerts.

Question 48

To save money, a manager wants to cut the database replicas for your core service from three to one. What do you do?

Accepted Answer

Show the savings next to the outage and data loss risk, and offer a cheaper, safer change.

Question 49

You find that a large cost overrun came from a batch job one engineer launched with the wrong setting. How do you handle the follow-up?

Accepted Answer

Treat it as a missing guardrail, and add limits and alerts for such jobs.

Question 50

Leadership asks whether your platform team should spend a quarter on cost work instead of features. How do you frame the answer?

Accepted Answer

Give the expected monthly savings, the effort, and which features would slip, so leaders can weigh them.

Question 51

A strong engineer on your team ships features quickly, but three times this quarter their changes broke other teams' integrations. What do you do first?

Accepted Answer

Talk with them privately, show the specific breakages, and ask how they see it.

Question 52

In that conversation, the fast engineer says contract tests slow them down and other teams should handle changes better. What do you say?

Accepted Answer

Agree speed matters, but say callers cannot adapt to unseen changes, and set the bar.

Question 53

How can you keep the fast engineer's speed while protecting contracts across the whole team?

Accepted Answer

Add consumer contract tests to continuous integration, plus safe fast paths such as additive changes.

Question 54

A month later, the fast engineer breaks another contract after agreeing to the new expectations. What now?

Accepted Answer

Name the gap plainly, call it a performance concern, and involve their manager.

Question 55

The fast-shipping engineer wants to be promoted to senior. What feedback best helps them?

Accepted Answer

Senior scope means others can rely on their work; give them a project with many callers to prove it.

Question 56

Another team's lead messages you angrily that your engineer broke their checkout flow again. How do you respond?

Accepted Answer

Own it, help restore service, and say what you are changing to prevent repeats.

Question 57

Your team publicly rewards whoever ships the most features each sprint. What is the best change you can push for?

Accepted Answer

Recognize the impact of shipped work, counting incidents and rework, instead of feature count.

Question 58

A new engineer starts copying the fast engineer's habit of skipping contract checks. What is the most effective way to set the norm?

Accepted Answer

Make the checks run by default in the build, so the right path is also the easy one.

Question 59

You are proposing a new event bus that eight teams will publish to. How should your design doc open?

Accepted Answer

The problem, who it affects, goals and non-goals, and the decision you need.

Question 60

In a design review for your caching layer, a senior engineer from another team strongly objects to your consistency model. What do you do?

Accepted Answer

Ask for a concrete failure case, and check whether your design handles it.

Question 61

Your design doc for a new job scheduler has been in review for six weeks and has 200 comments. What do you do?

Accepted Answer

Set a decision date, list the open questions, settle them with the key people, and close.

Question 62

Who should you invite to review a design that changes how all services authenticate to each other?

Accepted Answer

Security, owners of the most affected services, and the on-call leads who will run it.

Question 63

A reviewer leaves dozens of style comments on your design doc but says nothing about the core tradeoff. What do you do?

Accepted Answer

Fix what is cheap, and ask them directly about the tradeoff you most need checked.

Question 64

An engineer on your team writes a technically sound design doc for a billing change, but nobody reads it. What feedback helps most?

Accepted Answer

Lead with the decision and main tradeoff, and move the details to an appendix.