CI failure in "compute (re)moved files without deprecation"

Stream: mathlib4

[Michael Rothgang]

I just noticed #29544 fails in the post-build step with a surprising (to me) error:
https://github.com/leanprover-community/mathlib4/actions/runs/17639352125/job/50122548735?pr=29544 step.
To me, this smells like a bug in the "check out the right branches to compare" dance.
@Damiano Testa Would you like to take a look (no pressure!)? Thank you!