a wide refactor of aprender-contracts made three unrelated PRs DIRTY on the same three files — the class needs a rebase sweep the moment such a merge lands #3628

Measured in the 2026-09-20 sweep of DIRTY PRs. Three unrelated PRs — #3351 (pv L2 column), #3395 (AutoGluon CRUX category), #3350 (GDN shape keys) — were all cut on 2026-09-16 and were all DIRTY by 2026-09-18 on the same files under crates/aprender-contracts: src/lint/strict_test_binding.rs, src/commands/lint.rs, src/schema/types.rs / src/schema/validator.rs. One merge did it: ONT-6 (#3463, "one Verdict lattice, Kani-proved; per-repo arming; exit-vocabulary mapping"), a wide refactor of the contracts crate that landed on 2026-09-18. None of the three PRs is about the Verdict lattice; each touched those files for its own reason and each now carries a content conflict the by-id roadmap resolver cannot help with, because the conflicts are in code. All three are otherwise green (present is their only red) and none is superseded. The class: a wide refactor of a shared crate makes every open PR that touches that crate DIRTY at once, silently, and the conflicts surface one PR at a time as each is picked up — days later, by someone who did not land the refactor. No proposal here; this is a ledger entry for the failure-classes list.

Refs #3463, #3351, #3395, #3350