diff --git a/ChangeLog.md b/ChangeLog.md
index e5c8ef3..61c008d 100644
--- a/ChangeLog.md
+++ b/ChangeLog.md
@@ -8,9 +8,7 @@ Generated by [`auto-changelog`](https://github.com/CookPete/auto-changelog).
 
 ##### Bug Fixes
 
-- Add tests for APRSDClient._checks initialisation and first-call keepalive behaviour [`20da2ad`](https://github.com/craigerl/aprsd/commit/20da2ad)
-
-- Fix APRSDClient._checks AttributeError on first keepalive_check() call — initialise to False in __init__ [`52e1346`](https://github.com/craigerl/aprsd/commit/52e1346)
+- Fix WatchList.stats() and add regression tests — remove early return that bypassed lock [`481a78a`](https://github.com/craigerl/aprsd/commit/481a78a)
 
 - Fix APRSISDriver.is_configured() always returning True when driver is disabled [`a9ef65f`](https://github.com/craigerl/aprsd/commit/a9ef65f)
 
diff --git a/aprsd/packets/watch_list.py b/aprsd/packets/watch_list.py
index ec09128..e0bb1e1 100644
--- a/aprsd/packets/watch_list.py
+++ b/aprsd/packets/watch_list.py
@@ -51,7 +51,6 @@ class WatchList(objectstore.ObjectStoreMixin):
     @trace.no_trace
     def stats(self, serializable=False) -> dict:
         stats = {}
-        return self.data
         with self.lock:
             for callsign in self.data:
                 stats[callsign] = {
diff --git a/tests/packets/test_watch_list.py b/tests/packets/test_watch_list.py
index 9010d9b..c2d5ef0 100644
--- a/tests/packets/test_watch_list.py
+++ b/tests/packets/test_watch_list.py
@@ -14,8 +14,10 @@ class TestWatchList(unittest.TestCase):
 
     def setUp(self):
         """Set up test fixtures."""
-        # Reset singleton instance
+        # Reset singleton AND class-level state fully between tests
         watch_list.WatchList._instance = None
+        watch_list.WatchList.data = {}
+        watch_list.WatchList.initialized = False
         # Mock config
         CONF.watch_list.enabled = True
         CONF.watch_list.callsigns = ['TEST*']
@@ -24,6 +26,8 @@ class TestWatchList(unittest.TestCase):
     def tearDown(self):
         """Clean up after tests."""
         watch_list.WatchList._instance = None
+        watch_list.WatchList.data = {}
+        watch_list.WatchList.initialized = False
 
     def test_singleton_pattern(self):
         """Test that WatchList is a singleton."""
@@ -47,6 +51,58 @@ class TestWatchList(unittest.TestCase):
         self.assertIn('TEST1', wl.data)
         self.assertIn('TEST2', wl.data)
 
+    def test_stats_empty(self):
+        """stats() returns an empty dict when the watch list is empty."""
+        watch_list.WatchList._instance = None
+        CONF.watch_list.callsigns = []
+        wl = watch_list.WatchList()
+        self.assertEqual(wl.stats(), {})
+
+    def test_stats_returns_enriched_shape(self):
+        """stats() must return the enriched dict shape, not the raw internal data.
+
+        Regression test for the early-return bug: 'return self.data' was the
+        first statement in stats(), bypassing the lock and returning the raw
+        internal dict instead of the expected {callsign: {last, packet, age, old}}
+        shape that callers rely on.
+        """
+        watch_list.WatchList._instance = None
+        CONF.watch_list.callsigns = ['ENRICHED*']
+        wl = watch_list.WatchList()
+
+        # Populate with a seen packet so there is something to report
+        from tests import fake
+
+        packet = fake.fake_packet(fromcall='ENRICHED')
+        wl.rx(packet)
+
+        stats = wl.stats()
+
+        self.assertIn('ENRICHED', stats)
+        entry = stats['ENRICHED']
+        # These keys are built by the locked loop — absent if early return fires
+        self.assertIn('last', entry)
+        self.assertIn('packet', entry)
+        self.assertIn('age', entry)
+        self.assertIn('old', entry)
+        # Confirm we are NOT getting the raw internal key that only exists there
+        self.assertNotIn('was_old_before_update', entry)
+
+    def test_stats_not_raw_internal_dict(self):
+        """stats() must not return the raw self.data reference.
+
+        If stats() returns self.data directly, mutations via rx() after the
+        call would silently alter the returned value — and any consumer
+        that sees 'was_old_before_update' knows it got the raw dict.
+        """
+        watch_list.WatchList._instance = None
+        CONF.watch_list.callsigns = ['RAW*']
+        wl = watch_list.WatchList()
+
+        stats = wl.stats()
+        # stats() should return a freshly-built dict, not the live self.data ref
+        self.assertIsNot(stats, wl.data)
+
     def test_stats(self):
         """Test stats() method."""
         wl = watch_list.WatchList()