Stop a back-dated call from silently disabling every recency gate (server-26#74)

_call_fits_incident measured incident idle with the signed helper while every
other recency gate in the file uses the unsigned one. On the re-correlation
sweep, `now` is the call's own started_at, which can precede the incident's
last activity, so the value went negative.

Negative idle made `idle_min >= 15` false, which meant the content-divergence
veto never ran and unit overlap was accepted unconditionally -- on a shared
dispatch backbone that is the feedback loop that lets one incident absorb a
whole talkgroup. It also made `idle_min < 20.0` true at any back-dating, so a
tactical channel returned tactical_default for every swept orphan out to the
90-minute bound.

One variable feeds all four gates in the function, so this is a one-line change
at the source. The signed value is untouched where it belongs: callers still
compute corr_incident_idle_min themselves, so debug output keeps its meaning.

Direction is toward more splitting, on the sweep path only, which is the point
-- the bug was suppressing an over-merge veto. Forward-dated calls and anything
inside the thresholds behave exactly as before.

Two tests added alongside the existing idle-gate cases; both fail on the old
line and pass on the new one. 266 pass, 0 fail.

Refs server-26#74, #5, #80.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

drb-c2-core/app/internal/incident_correlator.py

@@ -1598,7 +1598,15 @@ def _call_fits_incident(
     Thin calls (no units/vehicles/coords) never reach this function —
     they are intercepted before it in correlate_call.
     """
-    idle_min = _incident_idle_minutes(inc, now) if now is not None else 9999.0
+    # Gate comparisons in this function must use the unsigned distance, not the
+    # signed value: the re-correlation sweep anchors `now` to the call's own
+    # started_at, which can be earlier than the incident's last activity and
+    # send the signed value negative — silently defeating every `idle_min`
+    # gate below (content-divergence veto and the tactical default alike).
+    # See `_idle_gate_minutes` docstring. The signed value is reported to
+    # callers separately as `corr_incident_idle_min` (they compute it via
+    # `_incident_idle_minutes` themselves) — nothing here needs it.
+    idle_min = _idle_gate_minutes(inc, now) if now is not None else 9999.0
     inc_id   = inc.get("incident_id", "?")
 
     # ── 1. Unit overlap ───────────────────────────────────────────────────────

drb-c2-core/tests/test_correlator_merge_caps.py

@@ -23,9 +23,10 @@ Three defects combined to produce it, and each has cases below:
 import pytest
 from datetime import datetime, timedelta, timezone
 from app.config import settings
+import app.internal.incident_correlator as correlator_mod
 from app.internal.incident_correlator import (
     _run_decision, _is_thin_call, _idle_gate_minutes,
-    _incident_at_capacity, _incident_span_minutes,
+    _incident_at_capacity, _incident_span_minutes, _call_fits_incident,
 )
 
 NOW = datetime(2026, 8, 20, 7, 0, 0, tzinfo=timezone.utc)
@@ -221,6 +222,49 @@ def test_back_dated_thin_call_does_not_sail_through_the_recency_gate():
     assert _run_decision(_ctx(all_active=[future], recent=[future]))["action"] == "orphan"
 
 
+def test_back_dated_call_does_not_bypass_the_content_divergence_veto(monkeypatch):
+    """
+    Same `9d376ffe` failure mode, but exercised directly against
+    `_call_fits_incident` on a dispatch channel: unit overlap plus a
+    back-dated call (incident updated 45 minutes AFTER the call's own
+    `started_at`, which the sweep passes as `now`) used to make the signed
+    idle -45, so `idle_min >= 15` read False and the content-divergence
+    veto never ran — unit overlap alone forced the merge regardless of
+    what the call was actually about. With the gate fixed to compare
+    distance, idle_min is 45 (>= 15), the veto runs, and a divergent
+    embedding (patched below so the assertion doesn't depend on numpy
+    being installed in this environment) fails it.
+    """
+    monkeypatch.setattr(correlator_mod, "_cosine_similarity", lambda a, b: 0.0)
+    inc = _incident(idle_minutes=-45, units=["6-Adam"])
+    inc["embedding"] = [1.0, 0.0]
+    fits, signal = _call_fits_incident(
+        inc, call_units=["6-Adam"], call_vehicles=[], call_coords=None,
+        proximity_km=settings.location_proximity_km, is_dispatch=True,
+        call_embedding=[0.0, 1.0], now=NOW,
+    )
+    assert (fits, signal) == (False, "content_divergence")
+
+
+def test_back_dated_call_on_tactical_channel_does_not_get_the_default():
+    """
+    Tactical-channel counterpart: no unit/vehicle/location signal, so the
+    function falls through to step 4's `idle_min < 20.0` default. A
+    back-dated call (incident updated 45 minutes after the call's own
+    started_at) used to read idle_min as -45, which is always < 20.0, so
+    `tactical_default` fired unconditionally no matter how stale the
+    incident actually was relative to this call. Fixed, idle_min is the
+    45-minute distance, which is not < 20.0.
+    """
+    inc = _incident(idle_minutes=-45)
+    fits, signal = _call_fits_incident(
+        inc, call_units=[], call_vehicles=[], call_coords=None,
+        proximity_km=settings.location_proximity_km, is_dispatch=False,
+        call_embedding=None, now=NOW,
+    )
+    assert (fits, signal) == (False, "tactical_idle")
+
+
 # ---------------------------------------------------------------------------
 # 4. Hard caps — path-independent, because pairwise fit tests can't see shape
 # ---------------------------------------------------------------------------

infra/firestore/firestore.indexes.json

@@ -30,6 +30,15 @@
         { "fieldPath": "ended_at", "order": "ASCENDING" }
       ]
     },
+    {
+      "//": "c2-core internal/dedup.py:84 — system_id == X AND started_at within a +/- window. Was live-failing on essentially every inbound call (server-26#84): dedup caught the FAILED_PRECONDITION and degraded to \"not a duplicate\", so double-heard transmissions were stored twice and would have been transcribed and correlated twice the moment an AI window opened. Created directly on c2-server 2026-08-28.",
+      "collectionGroup": "calls",
+      "queryScope": "COLLECTION",
+      "fields": [
+        { "fieldPath": "system_id", "order": "ASCENDING" },
+        { "fieldPath": "started_at", "order": "ASCENDING" }
+      ]
+    },
     {
       "//": "c2-core internal/vocabulary_learner.py:290 — system_id == X AND ended_at >= cutoff. Backend only; was live but undeclared until 2026-08-23.",
       "collectionGroup": "calls",