Skip to content

Defect Management

mavrovde edited this page Nov 1, 2026 · 1 revision

Defect Management

How defects (ISTQB: a flaw in a work product that can cause a failure) are reported, classified, fixed and learned from. Part of the Testing Strategy. Security vulnerabilities reported privately follow SECURITY.md and the deadlines of SSC-015 on Spec: Secure SDLC and Supply Chain; production incidents follow Spec: Operations. This page covers everything else and the test side of both.

1. Severity and priority

Severity is the impact of the defect; priority is the order in which it is worked on. Severity is set by the reporter and confirmed in triage; priority is set in triage.

Severity Meaning Examples
S1 Critical a licence, security or privacy guarantee is broken, or the service is unusable an item served to a plan its licence forbids; a key in a log; a request running past 2 s without cancellation
S2 Major a documented behaviour is wrong for a whole operation or class of requests, without a workaround wrong problem type for a whole operation; parity failure on a feature; budget breached by a wide margin
S3 Minor wrong behaviour with a workaround, or in an edge case a localisation missing in one language; a header missing on one status
S4 Trivial cosmetic or documentation wording a typo in a problem title
Severity Default priority Label Release impact
S1 P1 priority: P1 blocks every release until fixed
S2 P1 or P2 priority: P1 or priority: P2 blocks the release that would ship it
S3 P2 or P3 priority: P2 or priority: P3 listed in the completion report
S4 P3 priority: P3 —

Every defect issue carries type: bug, one area: … label, a priority label and a milestone. The severity is written in the first line of the issue body (Severity: S2), so it survives label changes. Flaky tests are defects of the test (type: bug, area: testing).

2. Defect report template

Severity: <S1-S4>

## What happened
<observed behaviour, with the request and response, keys redacted>

## What should happen
<expected behaviour, with the requirement IDs>

## How to reproduce
<environment, image digest or commit, fixture profile and seed, steps>

## Evidence
<test name, CI run link, log lines with secrets redacted>

## Level that should have caught it
<static | component | component integration | integration | contract | system | non-functional | acceptance>

A report never contains a real key, a secret or a full client address.

3. Lifecycle

stateDiagram-v2
  [*] --> New
  New --> Triaged: severity, priority, labels, milestone
  New --> Rejected: not a defect or duplicate
  Triaged --> InProgress: assigned
  InProgress --> InReview: fix and regression test in one pull request
  InReview --> Closed: merged with green gates
  Closed --> Reopened: failure seen again
  Reopened --> InProgress
  Rejected --> [*]
  Closed --> [*]
Loading

Text description of the diagram: a new defect is triaged or rejected; a triaged defect is assigned and worked on; the fix and its regression test go into one pull request; after review and green gates it is closed; if the failure reappears it is reopened.

Triage targets (design): S1 within one working day, S2 within three, others within the next weekly triage.

4. Regression rule

  • Every fix adds a test that fails without the fix (TST-003); the pull request records the red run (TST-013).
  • The test sits at the lowest level able to catch the defect (TST-020). If a higher-level test found it, the fix also asks why the lower level did not, and adds the lower-level test where it can.
  • The test carries the requirement ID it protects; if no requirement covered the behaviour, the specification gap is raised (TEST-012).

5. Escaped defects

A defect found after a release is an escaped defect. For each one, the completion report of the next release records the level that should have caught it and the change made to tests, data or gates. Escaped defects and defect removal efficiency are tracked as metrics on the Testing Strategy section 11.

6. Root cause

S1 and S2 defects get a short blameless root-cause note in the issue before closing: the cause, why the tests did not catch it, and what changed. Security S1 defects also follow the postmortem rules of Spec: Operations.

Home

About

Design

Data Model

Technical specification

Testing

Engineering

Technology choice

Reference

Clone this wiki locally