fix(mcp): a failing router in get_overview says since when, as the page does (#5860)

router_findings never passed `since`, so every failing router read as
brand new and sorted last among the criticals. The overview now reads
the failing routers' last day of history and dates each one with
problem_since, as the dashboard widget does.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

plugins/traefik/mcp_tools.py

@@ -14,6 +14,7 @@ from sqlalchemy import select
 
 from steward.core import anomalies as anom
 from steward.core.norms import norms_at
+from steward.core.time_range import bucket_seconds
 from steward.mcp._context import mcp_app
 from steward.mcp.tools._shape import (
     age_s, bounded, clamp_limit, db_session, iso, lookback, now_utc, rnd, summarize,
@@ -279,7 +280,8 @@ def cert_findings(certs, now, thresholds: dict) -> list[dict]:
 def router_findings(routers: list[dict]) -> list[dict]:
     """A router answering 5xx past its cutoff: failing. 4xx and latency are
     left to the metric anomalies, which judge them against the router's
-    normal (System 182)."""
+    normal (System 182). `since` is when the failing began, as the page's
+    verdict says it (r["since"], None when too recent for a bucket to show)."""
     out = []
     for r in routers:
         if r["state"] != "crit":
@@ -287,14 +289,29 @@ def router_findings(routers: list[dict]) -> list[dict]:
         said = (r.get("meaning") or {}).get("meaning")
         out.append(finding("critical", "traefik_router_failing", r["name"],
                            f"Traefik router {r['name']}: {r['why']}" + (f". {said}" if said else ""),
-                           ids={"router": r["name"]}, next=f'get_router(router="{r["name"]}")'))
+                           since=r.get("since"), ids={"router": r["name"]},
+                           next=f'get_router(router="{r["name"]}")'))
     return out
 
 
+async def failing_since(db, routers: list[dict], now, thresholds: dict) -> None:
+    """Sets r["since"] on each failing router from its last day of history,
+    as the dashboard widget reads it (#5860)."""
+    failing = [r["name"] for r in routers if r["state"] == "crit"]
+    if not failing:
+        return
+    since = now - page.WIDGET_SINCE
+    history = await page.router_history(db, failing, since, bucket_seconds(since))
+    for r in routers:
+        if r["name"] in history:
+            r["since"], _before = page.problem_since(history[r["name"]], thresholds)
+
+
 async def mcp_overview(db) -> list[dict]:
     thresholds, scrape = _cfg()
     now = now_utc()
     current, _gone = await _graded(db, now, thresholds, scrape)
+    await failing_since(db, current, now, thresholds)
     served, _old = page.current_certs((await db.execute(select(TraefikCert))).scalars().all(), scrape)
     return router_findings(current) + cert_findings(served, now, thresholds)
 

tests/plugins/traefik/test_mcp_tools.py

@@ -37,6 +37,14 @@ def test_only_a_failing_router_is_an_overview_finding():
     [f] = t.router_findings(routers)
     assert f["severity"] == "critical" and f["subject"] == "app@swarm"
     assert "Mostly 502" in f["summary"] and f["next"] == 'get_router(router="app@swarm")'
+    assert f["since"] is None
+
+
+def test_a_failing_router_says_since_when():
+    began = NOW - timedelta(hours=2)
+    [f] = t.router_findings([{"name": "app@swarm", "state": "crit", "why": "12.0% of responses are 5xx",
+                              "since": began}])
+    assert f["since"] == "2026-10-15T10:00:00Z" and f["age_s"] > 0
 
 
 def test_a_router_is_found_with_or_without_its_provider():