diff --git a/changelog.d/8971.bugfix b/changelog.d/8971.bugfix
new file mode 100644
index 0000000000..c3e44b8c0b
--- /dev/null
+++ b/changelog.d/8971.bugfix
@@ -0,0 +1 @@
+Fix small bug in v2 state resolution algorithm, which could also cause performance issues for rooms with large numbers of power levels.
diff --git a/synapse/state/v2.py b/synapse/state/v2.py
index f85124bf81..e585954bd8 100644
--- a/synapse/state/v2.py
+++ b/synapse/state/v2.py
@@ -658,7 +658,7 @@ async def _get_mainline_depth_for_event(
     # We do an iterative search, replacing `event with the power level in its
     # auth events (if any)
     while tmp_event:
-        depth = mainline_map.get(event.event_id)
+        depth = mainline_map.get(tmp_event.event_id)
         if depth is not None:
             return depth
 
diff --git a/tests/state/test_v2.py b/tests/state/test_v2.py
index 09f4f32a02..77c72834f2 100644
--- a/tests/state/test_v2.py
+++ b/tests/state/test_v2.py
@@ -88,7 +88,7 @@ class FakeEvent:
         event_dict = {
             "auth_events": [(a, {}) for a in auth_events],
             "prev_events": [(p, {}) for p in prev_events],
-            "event_id": self.node_id,
+            "event_id": self.event_id,
             "sender": self.sender,
             "type": self.type,
             "content": self.content,
@@ -381,6 +381,61 @@ class StateTestCase(unittest.TestCase):
 
         self.do_check(events, edges, expected_state_ids)
 
+    def test_mainline_sort(self):
+        """Tests that the mainline ordering works correctly.
+        """
+
+        events = [
+            FakeEvent(
+                id="T1", sender=ALICE, type=EventTypes.Topic, state_key="", content={}
+            ),
+            FakeEvent(
+                id="PA1",
+                sender=ALICE,
+                type=EventTypes.PowerLevels,
+                state_key="",
+                content={"users": {ALICE: 100, BOB: 50}},
+            ),
+            FakeEvent(
+                id="T2", sender=ALICE, type=EventTypes.Topic, state_key="", content={}
+            ),
+            FakeEvent(
+                id="PA2",
+                sender=ALICE,
+                type=EventTypes.PowerLevels,
+                state_key="",
+                content={
+                    "users": {ALICE: 100, BOB: 50},
+                    "events": {EventTypes.PowerLevels: 100},
+                },
+            ),
+            FakeEvent(
+                id="PB",
+                sender=BOB,
+                type=EventTypes.PowerLevels,
+                state_key="",
+                content={"users": {ALICE: 100, BOB: 50}},
+            ),
+            FakeEvent(
+                id="T3", sender=BOB, type=EventTypes.Topic, state_key="", content={}
+            ),
+            FakeEvent(
+                id="T4", sender=ALICE, type=EventTypes.Topic, state_key="", content={}
+            ),
+        ]
+
+        edges = [
+            ["END", "T3", "PA2", "T2", "PA1", "T1", "START"],
+            ["END", "T4", "PB", "PA1"],
+        ]
+
+        # We expect T3 to be picked as the other topics are pointing at older
+        # power levels. Note that without mainline ordering we'd pick T4 due to
+        # it being sent *after* T3.
+        expected_state_ids = ["T3", "PA2"]
+
+        self.do_check(events, edges, expected_state_ids)
+
     def do_check(self, events, edges, expected_state_ids):
         """Take a list of events and edges and calculate the state of the
         graph at END, and asserts it matches `expected_state_ids`