Release hygiene: one or more deploy gates are failing

[github-actions]

main and development have diverged

main is not an ancestor of development: 1 commit(s) on main are
missing from development.

10c4d24 Merge pull request #480 from views-platform/development

Fix with the back-merge, as a merge commit — never a squash, which creates yet another
commit that is not main, so this would reopen:

gh pr create --base development --head main --title "chore: sync main back into development"

Every release promotion leaves one merge commit on main that development has not
seen. Expected immediately after a release; a problem if it persists.
See docs/guides/publishing_to_pypi.md §C.5 and #402.

A deploy gate failed

            "See publishing_to_pypi.md §C.5. Detection also runs daily "
            "via .github/workflows/release-topology.yml."
        )
E       AssertionError: main is not an ancestor of development. Normal right after a release, and healed by the back-merge: gh pr create --base development --head main, merged as a MERGE COMMIT (a squash recreates the divergence). See publishing_to_pypi.md §C.5. Detection also runs daily via .github/workflows/release-topology.yml.
E       assert 1 == 0
E        +  where 1 = CompletedProcess(args=['git', 'merge-base', '--is-ancestor', 'main', 'development'], returncode=1, stdout=b'', stderr=b'').returncode

tests/test_falsification_deploy_v160.py:159: AssertionError
____________ TestDF1MergeTopology.test_main_ancestor_of_development ____________

self = <test_falsification_deploy_v160_r2.TestDF1MergeTopology object at 0x7f181c7de9d0>

    def test_main_ancestor_of_development(self) -> None:
        result = subprocess.run(
            ["git", "merge-base", "--is-ancestor", "main", "development"],
            capture_output=True,
        )
        # 0 = ancestor, 1 = not; other codes mean refs/history missing
        # (CI shallow checkout) — local pre-deploy gate, skip there (C-320).
        if result.returncode not in (0, 1):
            pytest.skip(
                "main/development refs unavailable (shallow CI "
                "checkout) — deploy gate runs locally"
            )
>       assert result.returncode == 0, (
            "main is not an ancestor of development. Expected right "
            "after a release: promotion leaves a merge commit on main "
            "that development has not seen. Fix with the back-merge — "
            "gh pr create --base development --head main — merged as a "
            "MERGE COMMIT, never a squash (a squash creates yet another "
            "commit that is not main). See publishing_to_pypi.md §C.5."
        )
E       AssertionError: main is not an ancestor of development. Expected right after a release: promotion leaves a merge commit on main that development has not seen. Fix with the back-merge — gh pr create --base development --head main — merged as a MERGE COMMIT, never a squash (a squash creates yet another commit that is not main). See publishing_to_pypi.md §C.5.
E       assert 1 == 0
E        +  where 1 = CompletedProcess(args=['git', 'merge-base', '--is-ancestor', 'main', 'development'], returncode=1, stdout=b'', stderr=b'').returncode

tests/test_falsification_deploy_v160_r2.py:50: AssertionError
=========================== short test summary info ============================
SKIPPED [1] tests/test_falsification_deploy_v160.py:275: local-clone hygiene is not answerable on a fresh CI runner — actions/checkout leaves one local branch, so this would pass without inspecting anything. The remote half of this class DOES run in CI.
2 failed, 3 passed, 1 skipped in 1.97s

These gates used to run only where someone typed pytest (C-341). Reproduce locally with:

uv run pytest tests/test_falsification_deploy_v160.py tests/test_falsification_deploy_v160_r2.py

Note the local run checks more than this one: the local-clone branch hygiene gate skips
on CI by design, because a fresh runner has one local branch and would pass without
inspecting anything.

Detected by release-hygiene. This issue is reused rather than duplicated,
and closes itself once every check passes again.

[github-actions]

Still failing as of 2026-08-26T08:37Z — diverged=true, gates_failed=true. run

[Polichinel]

Healed. The back-merge landed as #482 at 09:01Z, merged as a merge commit.

git merge-base --is-ancestor main development now returns 0 on both the remote refs and locally, so the gate this issue reports passes.

This was the expected post-release divergence the issue itself describes — "one merge commit on main that development has not seen… expected immediately after a release; a problem if it persists." It persisted for about twenty minutes.

Worth noting the detection worked exactly as designed: the release-topology workflow filed this during the release, with the fix command in the body, before anyone had to notice the branches had drifted.