REQ-ACCEPTANCE-004: persist Phase-6 price and post-MAIN expectation into WorldState

[zendev-author]

Closes #440

Achieved outcome

The Phase-6 -> Phase-8 dispatch pipeline now persists its per-tick output into the next tick's authoritative WorldState.markets instead of relying on test-owned cross-tick memory. applyMarketStateTransition(world, context) is the explicit production boundary a caller applies between executeTick() calls to carry forward the tick's Phase-6 price and the post-Phase-8-MAIN MarketExpectationState observation (Handoff/04 sections 9 and 11). Phase-8's MAIN-pass effectiveDemand/offered/cleared aggregates are now computed unconditionally in TickContext.marketClearingAggregates, no longer gated behind the collectTelemetry toggle, so this authoritative transition never depends on the non-authoritative telemetry flag (REQ-MARKET-005 neutrality).

Tested revision

28af61e07628a9ce220ae2712b59e09daed0e001

Changed artifacts

• src/simulation/marketStateTransition.ts (new): applyMarketStateTransition(world, context) — updates world.markets (priceByGood/expectationsByGood) for every market/good touched this tick; throws rather than inventing a market/good endpoint that doesn't already exist on WorldState.
• src/simulation/tickOrchestrator.ts: adds TickContext.marketClearingAggregates (new MarketClearingAggregate type) and initializes it empty per tick.
• src/simulation/phase8MainMarketClearing.ts: hoists the MAIN-pass effectiveDemand/offered/cleared aggregate computation out from behind the collectTelemetry gate so it always populates context.marketClearingAggregates; telemetry building itself is unchanged.
• src/simulation/phase6MarketPriceFormation.ts: exports ZERO_EXPECTATION and a new parseMarketPriceKey() (inverse of marketPriceKey()) for reuse by the transition boundary.
• src/simulation/acceptance-004-m3-golden-gate.test.ts: rewrites the MTFX-I5 multi-tick dispatch test to seed a market into WorldState and persist state via applyMarketStateTransition() between real executeTick() calls instead of getFixtureCurrentPrice/getFixtureExpectation closures; adjusts the trailing Phase-8-only negative control to run against the pre-loop initialWorldState (since the final worldState now legitimately carries the persisted price); adds a new negative-control test proving that skipping applyMarketStateTransition() between ticks leaves price and expectation observably stuck.
• docs/spec/implementation_status.csv / docs/spec/IMPLEMENTATION_STATUS.md: updates the REQ-ACCEPTANCE-004 ledger row's PR field to this PR and appends evidence for this fix (see acceptance criterion 6). A first pass of this edit accidentally converted the whole CSV from LF to CRLF line endings (Python csv.writer's default lineterminator); a follow-up commit on this branch rewrote it with an explicit \n terminator, so the cumulative diff against master is confined to the one updated row.

Acceptance criteria

• 1. A Phase-6 price produced in tick N becomes the authoritative LocalMarket price carried into tick N+1; the test does not provide that price through getFixtureCurrentPrice. Verified: the closure is removed; the market's starting price is seeded once into WorldState.markets, and applyMarketStateTransition() carries each tick's Phase-6 output forward.
• 2. The informative Phase-8 MAIN observation in tick N updates canonical MarketExpectationState, and Phase 6 in tick N+1 reads that persisted lagged state without a test-owned expectation callback. Verified: getFixtureExpectation is removed; the test asserts worldState.markets.get(marketId)!.expectationsByGood.get(goodId)!.observationCount increments by exactly 1 every tick.
• 3. Phase 8 in each tested tick settles/clears at exactly that tick's Phase-6 price. Preserved unchanged from the prior revision.
• 4. Repricing occurs at most once per market/good per tick and remains within the configured bounds already required by MTFX-I5. Preserved unchanged (phase6InvocationCount structural proof, bounds assertions).
• 5. A controlled negative variant with the persistence transition removed or bypassed makes the focused multi-tick regression fail. New test: "without applying the production persistence boundary between ticks, Phase 6 repeats the same stale price/expectation instead of progressing" — proves price and expectationsByGood stay stuck when applyMarketStateTransition() is never called.
• 6. REQ-ACCEPTANCE-004 ledger status/evidence matches what is actually proven. Ledger row updated in this PR: stays IMPLEMENTED, evidence appended naming this PR and what it closes.
• 7. Required TypeScript and retained .NET checks pass on the repairing revision. See Checks below.

Checks

Check	Outcome	Evidence
npm run typecheck	passed	clean
npm test	passed	574/574 (40 files; 1 new test)
npm run build	passed	vite build succeeded
dotnet build --configuration Release --no-restore	passed	0 warnings, 0 errors
dotnet test --configuration Release --no-build	passed	45/45
python scripts/implementation_status.py --check	passed	after ledger update commit

Not checked

Nothing deliberately skipped for this Issue's scope.

Assumptions and unknowns

• The MTFX-I5-adjacent trailing negative control ("Phase-8 alone falls back to the world's price") previously depended on world.markets never being updated; now that persistence is real, that assertion had to move to run against the pre-loop initialWorldState rather than the post-loop worldState (which now legitimately carries the persisted final price). This is a necessary consequence of the fix, not a weakening — the new dedicated negative-control test covers the "persistence removed" case directly.
• This PR only wires the market/good pairs the M3 fixtures actually drive through context.marketPrices/context.marketClearingAggregates; it does not add a scheduled step that calls applyMarketStateTransition() automatically inside executeTick() or a higher-level run loop, because no such run loop exists yet in this codebase (M3 fixtures call executeTick() directly, tick by tick, from tests). Wiring an automatic per-tick caller is future run-loop work, not part of this Issue's scope.

Highest-risk area for review

applyMarketStateTransition()'s decision to throw when a touched marketId is not already present in world.markets, rather than inventing an entry. This matches the "fail rather than invent an endpoint" pattern used elsewhere in this codebase (e.g. Issue #427's settlement design), but means every market/good a Phase-6/Phase-8 fixture touches must be seeded into WorldState.markets ahead of time — worth confirming this is the right boundary before production Phase-2/3 planning replaces these fixtures.

Remaining gate

None identified for this Issue's scope.

[zendev-acceptor]

ACCEPT — head revision 28af61e07628a9ce220ae2712b59e09daed0e001.

1. Required checks green at head: build-and-test, typescript, policy-guard, mergeability all pass at this exact head (verified via gh pr checks 441); mergeStateStatus is CLEAN.
2. Acceptance criteria (Issue #440), each verified against the diff:
   • AC1/AC2: applyMarketStateTransition() (new src/simulation/marketStateTransition.ts) persists Phase-6 price and post-Phase-8-MAIN MarketExpectationState into WorldState.markets; the rewritten multi-tick test seeds the market directly into WorldState and calls the real boundary between executeTick() calls — no getFixtureCurrentPrice/getFixtureExpectation closures remain.
   • AC3/AC4: same-tick price-sharing and phase6InvocationCount === tick + 1 once-per-tick assertions are unchanged in the diff (confirmed by direct read of the test file at this head).
   • AC5: new test "without applying the production persistence boundary between ticks..." runs 3 ticks without calling applyMarketStateTransition() and asserts price/expectation stay stuck — a real negative control.
   • AC6: ledger row updated, PR field points to #441, MERGE_COMMIT correctly left blank pending merge (consistent with this repo's established follow-up-PR pattern for recording merge commits).
   • AC7: see check reproduction below.
3. Scope: diff confined to the 7 files declared in "Changed artifacts" (marketStateTransition.ts new, tickOrchestrator.ts, phase8MainMarketClearing.ts, phase6MarketPriceFormation.ts, acceptance-004-m3-golden-gate.test.ts, ledger CSV/MD) — matches Issue #440's Scope. No .github/workflows/**, AGENTS.md, or docs/zendev/** touched.
4. No invariant/test weakening: the one relocated assertion (trailing Phase-8-alone negative control now runs against pre-loop initialWorldState) is a strengthening, not a weakening — it adds a concrete toBe(initialPrice) check alongside the original not.toBe(lastPrice) check, justified by persistence now legitimately advancing worldState.
5. No secrets/credentials/personal data/local paths found in the diff.
6. Handoff record: all 9 required sections present in the PR body (outcome, tested revision, changed artifacts, acceptance criteria, checks, not-checked, assumptions/unknowns, highest-risk area, remaining gate).

Independent re-execution at this head (checked out via worktree, not trusting the checks tab):

• npm ci && npm run typecheck — clean
• npm test — 574/574 passed (40 files)
• npm run build — succeeded
• dotnet restore && dotnet build --configuration Release --no-restore — 0 warnings, 0 errors
• dotnet test --configuration Release --no-build — 45/45 passed
• python3 scripts/implementation_status.py --check — matches

No prior verdict exists on this PR (reviews list empty) and no standing refusal from any account. Merging.