All sessions invalidated: symptoms, cause, and the test that catches it
A change to session signing or storage logs every user out at once.
All sessions invalidated is documented by 3 accepted public reports from 3 organisations, 3 of them published in the last 5 years.
- Likelihood: medium (3 public reports in 5 years)
- Impact: medium (3 of 3 reports: availability)
Symptoms
User-visible
- Users get "Token invalid" errors and every logged-in user on the Android app and web admin must log in again.[2]
- Users in one region are logged out repeatedly while browsing the web UI, while API access keeps working.[1]
- Users see Session Expired errors, are forced out, and some cannot log in at all.[3]
- Session Expired errors keep appearing at a lower rate after the release is rolled back.[3]
- Every user is logged out after each deployment and has to log in again.[4]
Monitoring-visible
- Token validation errors surge but raise no alarm when those errors are suppressed in logging.[2]
- Customer reports of repeated logouts began about 26 minutes after the deployment started.[1]
- Volunteers and users reported Session Expired errors and forced logouts shortly after the deployment started.[3]
Root causes
- stage: config A hotfix for TLS handshake errors changed the secret key used to validate access tokens during a configuration clean-up, so every existing token failed validation.[2]
- stage: release-process A rolling deployment moved session storage to a new cache while instances updated at different times, so a session created on one instance was missing on the next.[1]
- stage: code A framework update switched session cookies to an encrypted format with a SHA256 digest, so old and new instances could not read each other's cookies during the rolling deployment.[3]
- stage: code An edge worker script that parses the session cookie could not read the new format and treated requests as signed in without a valid session.[3]
Public examples
- [1]Phrase, 2026-02-23. A deployment migrated session storage to a new caching solution while application instances were updated at different times, so instances used different session stores and UI users in the EU region were repeatedly logged out. primary Performance Disruption of Phrase TMS CAT Web Editor and Project Management (EU) on January 29, 2026 between 4:56 PM CET and 5:46 PM CET
- [2]Mews, 2025-04-11. A hotfix for a TLS issue changed a secret key during a configuration clean-up, which invalidated every logged-in session on Android and web admin and forced all users to log in again. primary Login issue on Mews POS app
- [3]Organization for Transformative Works, 2024-10-07. A new Archive of Our Own release changed session cookies to an encrypted format and switched cookie digests from SHA1 to SHA256 as part of a Rails update, which during a rolling deployment produced Session Expired errors and forced logouts. primary AO3 elevated "Session Expired" errors
Practitioner sources cited for symptoms
- [4]Can Olcer, 2021-10-29. A developer describes a Rails app that logged out all users after every deployment because the hosting platform set a new secret_key_base on each deploy, so existing session cookies could not be decrypted. practitioner Rails: Prevent users from logging out after each deployment
Test classes that catch it
| Test class | Stage | Limitation | Basis |
|---|---|---|---|
| Configuration and environment tests | pre-deploy | A configuration test that confirms the signing secret is present and unchanged misses a cookie format change made in code and a session store split between instances on different releases. | by design analysis |
| End-to-end tests | pre-deploy | An end-to-end login test with a fresh session passes, because the fault hits only sessions created before the deploy and shows only while old and new instances serve the same user. | by design analysis |
| Canary and progressive delivery checks | post-deploy | A canary flags the logout surge only if token validation errors reach logs and alarms, so suppressed validation errors leave it blind, and the mixed old and new fleet it creates is itself the condition that splits sessions. | by design analysis |
Risk
- Likelihood: medium (3 public reports in 5 years)
- Impact: medium (3 of 3 reports: availability)
| Accepted public reports | 3 |
|---|---|
| Published in the last 5 years | 3 |
| Impact stated: availability | 3 |
Pre-release checklist
- PR-01
- PR-02
- PR-03
- PR-04
- PR-05
- PR-06
- PR-07
First 30 minutes
- FM-01
- FM-02
- FM-03
- FM-04
- FM-05
- FM-06
Download the checklist
What catches this: catalogue entries
Listed because their own documentation or site names configuration and environment tests, end-to-end tests, canary and progressive delivery checks on the verification date. Alphabetical. Not a recommendation.
Tools whose documentation names these test classes
- Appium e2e source verified 2026-10-21
- Applitools Eyes e2e source verified 2026-10-21
- Argo Rollouts canary source verified 2026-10-21
- axe-core e2e source verified 2026-10-21
- behave e2e source verified 2026-10-21
- Checkly e2e source verified 2026-10-21
- CodeceptJS e2e source verified 2026-10-21
- Cucumber e2e source verified 2026-10-21
- Cypress e2e source verified 2026-10-21
- Datadog Synthetic Monitoring e2e source verified 2026-10-21
- Detox e2e source verified 2026-10-21
- Espresso e2e source verified 2026-10-21
- Firebase Test Lab e2e source verified 2026-10-21
- Flagger canary source verified 2026-10-21
- Flipt config source verified 2026-10-21
- GrowthBook config source verified 2026-10-21 canary source verified 2026-10-21
- Karate e2e source verified 2026-10-21
- Katalon Studio e2e source verified 2026-10-21
- Kayenta canary source verified 2026-10-21
- Kobiton e2e source verified 2026-10-21
- LaunchDarkly config source verified 2026-10-21 canary source verified 2026-10-21
- Maestro e2e source verified 2026-10-21
- Mocha e2e source verified 2026-10-21
- Nightwatch e2e source verified 2026-10-21
- Pa11y e2e source verified 2026-10-21
- Percy e2e source verified 2026-10-21
- Playwright e2e source verified 2026-10-21
- PostHog config source verified 2026-10-21
- Puppeteer e2e source verified 2026-10-21
- Robot Framework e2e source verified 2026-10-21
- RSpec e2e source verified 2026-10-21
- Sauce Labs e2e source verified 2026-10-21
- Selenium e2e source verified 2026-10-21
- Serenity BDD e2e source verified 2026-10-21
- Spinnaker canary source verified 2026-10-21
- TestCafe e2e source verified 2026-10-21
- TestingBot e2e source verified 2026-10-21
- TestMu AI (formerly LambdaTest) e2e source verified 2026-10-21
- Unleash config source verified 2026-10-21 canary source verified 2026-10-21
- WebdriverIO e2e source verified 2026-10-21
- XCTest e2e source verified 2026-10-21
Providers whose site names these test classes
- 2i e2e source verified 2026-10-21
- Abstracta e2e source verified 2026-10-21
- andagon people GmbH e2e source verified 2026-10-21
- Appmatics e2e source verified 2026-10-21
- AQA Masters e2e source verified 2026-10-21
- BugRaptors e2e source verified 2026-10-21
- Cleverix e2e source verified 2026-10-21
- Crestech Software Systems e2e source verified 2026-10-21
- Crowdar e2e source verified 2026-10-21
- DeviQA e2e source verified 2026-10-21
- Elinext e2e source verified 2026-10-21
- Exactpro e2e source verified 2026-10-21
- Full Scale e2e source verified 2026-10-21
- Inmetrics e2e source verified 2026-10-21
- Invensis Technologies e2e source verified 2026-10-21
- iScale Solutions e2e source verified 2026-10-21
- Itekako e2e source verified 2026-10-21
- iValuePlus Services e2e source verified 2026-10-21
- Nimble Approach e2e source verified 2026-10-21
- Outpost QA e2e source verified 2026-10-21
- PLATO e2e source verified 2026-10-21
- Polteq e2e source verified 2026-10-21
- Q-Centric GmbH e2e source verified 2026-10-21
- QA Madness e2e source verified 2026-10-21
- QA Mentor e2e source verified 2026-10-21
- QASource e2e source verified 2026-10-21
- QAwerk e2e source verified 2026-10-21
- Quality House e2e source verified 2026-10-21
- Quality Island e2e source verified 2026-10-21
- QualityKiosk Technologies e2e source verified 2026-10-21
- QualityLogic e2e source verified 2026-10-21
- Scalefocus e2e source verified 2026-10-21
- SHIFT ASIA e2e source verified 2026-10-21
- SimplyTest GmbH e2e source verified 2026-10-21
- Sofka Technologies e2e source verified 2026-10-21
- Soflab Technology e2e source verified 2026-10-21
- SQA - Software Quality Assurance S.A. e2e source verified 2026-10-21
- SunTec India e2e source verified 2026-10-21
- T2S Tecnologia e2e source verified 2026-10-21
- TestFort e2e source verified 2026-10-21
- Testhouse e2e source verified 2026-10-21
- Testing IT e2e source verified 2026-10-21
- TestMatick e2e source verified 2026-10-21
- Testrig Technologies e2e source verified 2026-10-21
- Testscenario e2e source verified 2026-10-21
- TestSolutions GmbH e2e source verified 2026-10-21
- THEY.dev e2e source verified 2026-10-21
- tiQtoQ e2e source verified 2026-10-21
- VTEST e2e source verified 2026-10-21
- WebDepend e2e source verified 2026-10-21
- Zoonou e2e source verified 2026-10-21
Only on this page
In report [1], all sessions invalidated was detected when customers began reporting repeated logouts about 26 minutes after the deployment started. The team then identified inconsistent session storage configuration across instances [1].
The impact ended when the deployment was completed on every instance rather than rolled back [1]. The report's follow-up classifies session storage configuration changes as high risk and treats them as planned maintenance with advance notice.
Related incidents
- Password reset link broken Password reset emails link to a wrong or expired route after a routing change.
- Flag enabled for wrong segment A feature flag targets the wrong users, region or percentage after a rule edit.
- SSO login broken Single sign-on fails after an identity provider, certificate or callback URL change.
- API key rotation breaks integrations A key rotation invalidates partner integrations that were not notified.
- Autoscaling policy misconfigured A scaling change fails to add capacity or scales to zero under load.
Questions
What causes all sessions invalidated?
The 3 primary reports name a changed signing secret [2], a session store moved mid-rollout [1] and a session cookie format change [3]. Each made existing sessions unreadable by at least some instances.
What are the symptoms of all sessions invalidated?
Users see Token invalid or Session Expired errors and must log in again, in one report repeatedly. Monitoring can stay quiet when token validation errors are suppressed [2].
Why did tests not catch all sessions invalidated?
The fault hits sessions created before the deploy, which a fresh login test never holds. One report states that the smaller staging environment did not show the problem [3].
Which test class catches all sessions invalidated, and what does it miss?
Configuration tests, end-to-end tests and canary checks are listed by design, with no cited basis. A configuration test on the signing secret misses a cookie format change made in code.
How long did it take to detect all sessions invalidated in the public reports?
One report measures about 26 minutes from the start of the deployment to the first customer reports [1]. The other 2 primary reports give no measured detection time.
Should a team roll back or fix forward when all sessions invalidated happens?
One report ended impact by completing the rollout [1], and another rolled back and redeployed without the cookie changes [3]. Checklist item FM-03 picks the direction that puts every instance on one session format and one store.
Release Incident Playbook (2026). All sessions invalidated: symptoms, cause, and the test that catches it, content version 1.0.0. https://releasefailures.com/incidents/all-sessions-invalidated/. Accessed 2026-10-21.