Field notes for QA & test-automation teams — how to know what a PR breaks before CI tells you.
The diff is in src/, the failure is in tests/, and nothing connects them at review time. Here's how to close that gap.
Separating automation into its own repo is good hygiene — and it severs the only signal that a frontend PR breaks a test.
Why CSS and XPath selectors rot, and how data-testid / data-cy hooks make a suite survive refactors.
Not every red test matters equally. Use historical failure rates to surface the tests most likely to fail on this change.