Daily-trend veto on M15 + lot-size / already-open lines (campaign 2026-09-30), and the 2026-10-01 whole-codebase bug review

Two bodies of work that touch the same files, so one commit.

Direction and setup-size campaign (2026-09-30 → 10-01, owner's RISK-FIRST rule,
written before the runs):
- SIGNAL: M15 daily-trend veto (factory override 7, xau_strategy.daily_trend_veto
  "both"). A finished setup against a clear daily trend (closed D1 bars, Sunday
  stubs dropped; close vs its 50-bar EMA AND vs the close 60 kept days earlier)
  is removed as the last step of the core. One definition
  (signal_aggregator.daily_trend_series) for live, the bar replay and the twin's
  host side; D1 is a required frame of the M15 scan; mirrored in
  cuda_signal_kernels.py K2. W3 824 / +205.12R / PF 1.73 / max DD 12.4R vs
  1105 / +195.79R / 1.49 / 20.9R; 2019 → 2026 1473 / +223.33R / 1.42 / 41.5R vs
  2091 / +192.53R / 1.24 / 64.8R; unseen holdout −4.62R vs −6.63R.
- FAILED, not shipped: the veto on M5 (W3 drawdown 10.1R → 15.2R) and on H1
  (−9.7R), buy-only and sell-only on M15, the wide-stop cap
  (xau_strategy.wide_stop_cap_atr stays 0.0, a research knob).
- New M15 pins: W2 dfc647ee…, OOS e87b6de1…, W3 ad3b0dbd…. M5 / H1 / H4 pins
  unchanged. --set xau_strategy.daily_trend_veto=off reproduces the previous M15
  pins byte for byte. Twin and reference script MD5-equal on every window;
  recorder and feature-export parity re-proved; base-rate table rebuilt.
- DISPLAY (books byte-identical): exposure.py — the lot size for
  trade_size.risk_per_trade (new config.json block, off by default, private to
  the owner) and the count of alerted same-direction plans still open
  (signals.alerted_at, read-only open_alerted_setups); the daily trend in words
  on Telegram, the CLI and the dashboard; signals.daily_trend logged.
- providers/oanda: a date window with no bars is an empty page, not an error
  (a D1 pad starting on a Saturday could not be fetched).
- Reports only, nothing shipped: seven context screens (no state nominated) and
  eight outside data sources (none adds a forecast to gold's own daily trend).
- Independent verification (report section 5.13): every figure reproduced on
  both scripts, scoring identical to the previous commit on all four
  timeframes, no look-ahead. Caveats recorded there: this is a risk filter, not
  a more accurate direction call; nine tenths of the 2019 → 2026 gain is 2022;
  the neighbour clause is thin (a simple 100-day average misses the R line by
  1.2R); "never adds a trade" is an observation, not a guarantee.

Bug review of the whole codebase (2026-10-01). Signal path clean. 17 defects
elsewhere, plus 4 found by an independent review of the fixes; all fixed, each
with a test; no trade book changed:
- telegram_bot: one _deliver_alert for the candle scan and the level break. The
  row is stamped alerted_at BEFORE the send, on a one-worker alert-db pool, and
  un-stamped when no chat received it or the task is cancelled mid-stamp, so
  two alerts at one candle close count each other and no SQLite write runs on
  the event loop. The 5-bar cooldown counts bars of the scan's own frame
  (== gpu_backtest._apply_cooldown; clock time released the 22:00 UTC bar after
  the settlement pause) and is restored from signals.db at start-up. The
  lot-size line goes to the owner's own chat only (a group the owner ran
  /autoscan in used to receive it). /status gains a news-model line and escapes
  failure details.
- signal_log: mark_alerted returns the stamp it wrote; unmark_alerted;
  last_alerted_bars (read-only).
- advisory_outcome: candle continuity is judged bar by bar inside the walk; a
  hole after a trade's exit no longer voids it. The stored-candle replay
  resolves 108 of 111 logged setups instead of 99, all equal to the live
  resolver's R.
- providers/oanda: an empty date window is accepted only when it is 5 days or
  shorter and no malformed candle was skipped; a longer one is retried, then
  refused, instead of being cached as history with a hole.
- news_engine: blank RSS titles skipped; one unreadable headline no longer
  costs the local model its whole batch (checked on the real bundle); model
  health recorded; no DeepSeek audit call in a process that does not archive;
  event-direction cache keyed per release; peek_week_events never fetches.
- llm_client: chat_json never raises and never logs the API key (requests
  quotes the Authorization value in InvalidHeader; error bodies are redacted).
- signal_aggregator: the scan's news block is guarded; the replay path
  (use_news=False) assigns exactly the values it did before.
- agent_panel: PANEL_PROMPT_VERSION v2.2 — same prompt text, corrected input
  (event timing over the cached calendar week, headlines newest first).
  scripts/panel_replay.py replays the logged payload per persona;
  scripts/label_scan_outcomes.py counts with an ISO cutoff.
- test_all.py: no ALTER TABLE … DROP COLUMN (the live box has SQLite 3.34);
  tests that write Parquet print a notice and return without pyarrow.

Verification after the last code change: suite 632 passed, 0 failed (also with
both Parquet engines blocked and FB2_BACKEND=cpu, the live box's view); the
five reference MD5s reproduce on both backtesters, 0 cache fallbacks; Module 37
green; static Python 3.9 / SQLite 3.34 check clean on every touched file.

Evidence: backtest/research/daily_trend_campaign_2026-09-30.md,
context_screens_2026-10.md, new_inputs_2026-10.md; CLAUDE.md sections 2, 4, 6,
8, 9, 10 and 11; pair_signals/xauusd.py docstring (M15 override 7).

providers/oanda.py

@@ -37,6 +37,10 @@ class OandaDataProvider(BaseDataProvider):
     RETRY_DELAYS_S = (2.0, 5.0, 10.0)
     RETRY_BUDGET_S = 45.0
     RETRY_DELAY_S = RETRY_DELAYS_S[0]  # first delay; kept for callers/tests
+    # The longest date window that may come back empty and still mean
+    # "nothing traded here": a weekend joined to a holiday. A longer window
+    # with no bars is an anomaly, not a closure (see _fetch_oanda_candles).
+    EMPTY_WINDOW_MAX = pd.Timedelta(days=5)
 
     OANDA_GRANULARITY = {
         Timeframe.M1: "M1",
@@ -96,6 +100,26 @@ class OandaDataProvider(BaseDataProvider):
                 symbol, timeframe, outputsize, start, end)
         return self._fetch_oanda_candles(symbol, timeframe, outputsize)
 
+    @staticmethod
+    def _empty_frame() -> pd.DataFrame:
+        """A normalized frame with no bars (a date window nothing traded in)."""
+        df = pd.DataFrame(
+            {col: pd.Series(dtype="float64")
+             for col in ("Open", "High", "Low", "Close", "Volume")},
+            index=pd.DatetimeIndex([], tz="UTC", name="Datetime"))
+        return df
+
+    @classmethod
+    def _closure_sized(cls, start: Optional[datetime],
+                       end: Optional[datetime]) -> bool:
+        """True for a date window short enough to be one scheduled closure."""
+        if start is None or end is None:
+            return False
+        try:
+            return pd.Timestamp(end) - pd.Timestamp(start) <= cls.EMPTY_WINDOW_MAX
+        except (TypeError, ValueError):
+            return False
+
     def _to_oanda_instrument(self, symbol: str) -> str:
         try:
             return self.OANDA_INSTRUMENT[symbol]
@@ -192,8 +216,25 @@ class OandaDataProvider(BaseDataProvider):
             raise OandaDataError(
                 f"Oanda candle response for {instrument} is not valid "
                 f"JSON: {e}") from e
+        # A DATE WINDOW with no bars in it (a weekend, a holiday, the closed
+        # days before a window's first session) is not a failure: fetch_range
+        # walks history in windows and counts an empty one as "nothing traded
+        # here". Raising for it made every backtest window whose padded start
+        # falls on a closed day unfetchable once fetch_range began retrying a
+        # failing window and aborting after three tries (2026-09-13); found
+        # 2026-09-30 on the D1 pad of a window starting on a Saturday. A
+        # `count` request (the live path) with nothing in it stays an error.
+        # Only a CLOSURE-SIZED window gets that reading (2026-10-01 review):
+        # an empty answer for a longer one (a 15-day M5 page, a 46-day M15
+        # page) cannot be a closure, and accepting it let fetch_range step
+        # over the window and cache history with a hole in it. Raising makes
+        # fetch_range retry the same window and refuse after three tries.
+        # (An M1 page is 3 days, so it would still pass; M1 is not fetched.)
+        closure_sized = self._closure_sized(start, end)
         candles = data.get("candles", [])
         if not candles:
+            if closure_sized:
+                return self._empty_frame()
             raise OandaDataError(
                 f"Empty candle response from Oanda for {instrument}")
 
@@ -233,6 +274,9 @@ class OandaDataProvider(BaseDataProvider):
                 "OHLC fields) — skipped", skipped_malformed, instrument)
 
         if not rows:
+            # Malformed candles mean the window DID trade: never "no bars".
+            if closure_sized and not skipped_malformed:
+                return self._empty_frame()
             raise OandaDataError(
                 f"No complete candles from Oanda for {instrument}")