What I was looking for
Why correct fixes kept failing the upstream test suites they were sent to.
What I found
Maintainers trust a green suite. When the pinned value is wrong, the suite argues against the fix, and a reviewer reasonably asks why the expected numbers changed.
My skfolio fix had to change four pinned values that encoded the start-is-a-peak bug. In ffn, a test added a month earlier pinned the unannualised Ulcer Performance Index, so the fix rewrote two tests to land.
The move
When a fix breaks a test, I ask where the expected value came from before I assume my fix is wrong.