fix: resolve 4 logic bugs found in code review

- utc_to_eastern: use zoneinfo instead of hardcoded EDT offset (-4)
  so start times are correct in both EST and EDT
- standings: fetch before truncate so a failed API call doesn't wipe
  existing standings data
- routes: call parse_games() once per request instead of three times
- scheduler: wrap run_pending() in try/except so an unhandled exception
  doesn't kill the background thread

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

tests/test_games.py

@@ -101,10 +101,16 @@ class TestGetGameOutcome:
 
 
 class TestUtcToEstTime:
-    def test_converts_utc_to_est(self):
+    def test_converts_utc_to_edt(self):
+        # April is EDT (UTC-4): 23:00 UTC → 07:00 PM EDT
         result = utc_to_eastern("2024-04-10T23:00:00Z")
         assert result == "07:00 PM"
 
+    def test_converts_utc_to_est(self):
+        # January is EST (UTC-5): 23:00 UTC → 06:00 PM EST
+        result = utc_to_eastern("2024-01-15T23:00:00Z")
+        assert result == "06:00 PM"
+
 
 class TestParseGames:
     def test_returns_empty_list_for_none(self):

tests/test_scheduler.py

@@ -39,3 +39,20 @@ class TestStartScheduler:
             start_scheduler()
 
         assert mock_schedule.run_pending.call_count >= 2
+
+    def test_continues_after_exception_in_run_pending(self, mocker):
+        mock_schedule = mocker.patch("app.scheduler.schedule")
+        call_count = {"n": 0}
+
+        def raise_then_stop(_):
+            call_count["n"] += 1
+            if call_count["n"] >= 2:
+                raise StopIteration
+
+        mock_schedule.run_pending.side_effect = RuntimeError("boom")
+        mocker.patch("app.scheduler.time.sleep", side_effect=raise_then_stop)
+
+        with pytest.raises(StopIteration):
+            start_scheduler()
+
+        assert mock_schedule.run_pending.call_count >= 2

tests/test_standings.py

@@ -209,12 +209,25 @@ class TestRefreshStandings:
         assert count == 2
 
     def test_does_not_insert_when_fetch_fails(self, mocker, tmp_path):
-        mocker.patch("app.standings.fetch_standings", return_value=None)
-        mocker.patch("app.standings.DB_PATH", str(tmp_path / "test.db"))
+        db_path = str(tmp_path / "test.db")
+        mocker.patch("app.standings.DB_PATH", db_path)
 
+        # Seed with existing data before the failed refresh
+        seed = [
+            {
+                "team_common_name": "Bruins",
+                "league_sequence": 1,
+                "league_l10_sequence": 2,
+            }
+        ]
+        mocker.patch("app.standings.fetch_standings", return_value=seed)
         refresh_standings()
 
-        conn = sqlite3.connect(str(tmp_path / "test.db"))
+        # Now simulate a fetch failure — existing data must be preserved
+        mocker.patch("app.standings.fetch_standings", return_value=None)
+        refresh_standings()
+
+        conn = sqlite3.connect(db_path)
         count = conn.execute("SELECT COUNT(*) FROM standings").fetchone()[0]
         conn.close()
-        assert count == 0
+        assert count == 1