Encode complete mapped sleep products in the shared filter, using stable
whole-frame witnesses and sorted first-occurrence owner/action identities.
Count only missing products as insertions, retain accepted VALUE marks, and
preserve source/alias validation without an exact visited-context table.
Enable all four built-in strategies in owned library/harness/CLI paths while
keeping the existing DPOR algorithms and CBS policy intact. Add independent
product shadows, real witness/resource artifacts and final release evidence.
B9 closure awaits this phase's merge; B10 and parked work stay excluded.
Co-authored-by: Codex <noreply@openai.com>
Summary
Add approximate symmetry for DPOR, completing the selected B9 implementation. CLI symmetry with an explicit BITSTATE store now supports all four strategies, including omitted
--strategy.Validation
Build, 1,907 tests, strict Javadoc, 185 browser checks, external Java 26 jar consumer. Fresh baseline/final default PIT 455/468 K (97.22%), all 468 identities/statuses unchanged; 94% floor retained. Shared context 102/102 K and final checked store 73/73 K. Evidence ledger distinguishes the wide census, final store run and inherited residuals/CLI coverage boundary.
CodeRabbit completed all 20 submitted files. Both Minor findings and the Trivial finding were patched and locally verified; no rerun for minor/trivial fixes.
Spec 91 records the contract. The budget covers membership, not total heap; no speedup or calibrated coverage guarantee is claimed. B9 remains in progress until K3 merges. B10 and parked roadmap work remain excluded.
Co-authored-by: Codex noreply@openai.com