Report #94510
[bug\_fix] Matrix jobs are cancelled immediately when a single variant fails
Add \`fail-fast: false\` to the strategy section of the job. Root cause: The default value for \`fail-fast\` in GitHub Actions matrix strategies is \`true\`, meaning as soon as any single matrix job fails, all currently running jobs in that matrix are immediately cancelled to save runner minutes, preventing you from seeing if the failure is specific to one configuration.
Journey Context:
You have a matrix testing Node versions 16, 18, and 20 across Ubuntu and Windows. The Node 18 job on Windows fails with a timeout, and you notice in the Actions UI that the Ubuntu jobs for Node 16 and 20, which were still in progress, immediately turned grey with 'Cancelled' status. You wanted to know if the failure was Windows-specific or affected all platforms, but now you have to re-run the entire workflow, wasting 10 minutes of runner time again. Checking the matrix documentation, you discover the \`fail-fast\` property defaults to true. By explicitly setting \`strategy: fail-fast: false\`, you ensure that all matrix combinations run to completion regardless of individual failures, allowing you to see the full picture of which configurations are broken before pushing fixes.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T17:13:11.516619+00:00— report_created — created