Re-verify: is tools/beads/import_github_golden_test.py's golden fixture actually broken by #1152? nf-iw4i ← Beads

Twinned from https://github.com/cfsh/ninefold/issues/1157 by tools/beads/import_github.py's reconcile pass.

The original report

A dev session running ./dev.py pr reported, verbatim: *"./dev.py pr's gate is currently red on main for a reason unrelated to my change: PR #1152 (merged 2026-08-20, another session, no longer running) broke tools/beads/import_github_golden_test.py's golden fixture."*

#1152 ("operationalize_design.py: mint the doc's per-task section into the task body") touched tools/nflib/bd.py, design_docs/tools/golden_test.py, design_docs/tools/operationalize_design.py, and — relevantly — tools/beads/test_data/011/bd_children_after.json, which is one of the fixture files tools/beads/import_github_golden_test.py reads.

What the office checked before filing (and could not reproduce)

At origin/main = d193620 (after #1152, and after #1148 merged cleanly on top of it), from a clean ~/nf/ninefold-ro checkout reset to origin/main:

``` $ python3 -m unittest tools.beads.import_github_golden_test -v test_real_drift_in_nf_tvc_is_corrected_and_1137_gets_a_new_twin ... ok Ran 1 test in 0.002s OK

$ python3 -m unittest design_docs.tools.golden_test -v ... 4 tests ... OK

$ ./presubmit.py --suite python Ran 2865 tests in 68.461s FAILED (failures=1, skipped=1) ```

The one python-suite failure is the known, pre-existing test_every_tests_directory_in_the_repo_carries_a_gdignore baseline issue (tools/tests missing .gdignore) — unrelated to golden fixtures or #1152. No open PR already fixes an import_github golden failure, and no commit after #1152 touched either golden test or its fixtures.

What this issue needs

Either: 1. Confirm it's genuinely gone — if a clean origin/main checkout on this box passes both golden tests and ./presubmit.py --suite python with only the known baseline failure, this was likely transient (a race against the shared bd workspace mid-merge, a stale local checkout in the reporting session's own worktree, or similar) and the issue can close as not reproducible with a note explaining what was checked. 2. Or find the real trigger — if it reproduces under ./dev.py pr's actual gate invocation specifically (not a bare unittest run), dig into what's different about that path (e.g. does it exercise the *real* shared bd workspace at ~/nf/beads rather than FakeBd? design 011 Amendment A1 moved to one shared workspace, not per-worktree copies — if the golden test's fixture drifted against live shared-workspace state rather than the checked-in JSON, that would explain a failure that a fixture-only unittest run can't see) and fix the actual root cause.

Either outcome is a valid close — this issue exists to get a definitive answer, not to presuppose the bug is real.

Test plan

./presubmit.py --suite python green (or explain in the PR/close comment exactly which pre-existing baseline failures remain, per this repo's "four failures are the expected baseline" note). If a real fix is needed, add a regression test that would have caught it under ./dev.py pr's actual gate path, not just a bare unittest invocation.