index_stepdown_unique.js fails intermittently in the no_passthrough_primary_driven_index_builds suite (seen on enterprise-windows-all-feature-flags-required). It passes locally and on most runs but occasionally fails on evergreen.

The failing assertion is IndexBuildTest.assertIndexes(primaryColl, 2, ["id", "a_1"]) — only id is present, so a_1 was never committed.

It seems there is a timing race between PDIB (primary driven index build) re-committing and abort-on-step-up.
The suite makes failure more likely via WTWriteConflictException / WTWriteConflictExceptionForReads failpoints (activationProbability 0.001), --shuffle, and slower Windows RBE hosts. Locally the build usually completes before the abort wins.

Here is a visualization of the specific failing case:

Temporary fix: re-add the tag `primary_driven_index_builds_incompatible_due_to_abort_on_step_up` that was removed as part of SERVER-127607.