diff --git a/plugins/platforms/matrix/adapter.py b/plugins/platforms/matrix/adapter.py
index ebe9ebbbf..dac4dbd16 100644
--- a/plugins/platforms/matrix/adapter.py
+++ b/plugins/platforms/matrix/adapter.py
@@ -2351,7 +2351,18 @@ class MatrixAdapter(BasePlatformAdapter):
         if inspect.isawaitable(tasks):
             tasks = await tasks
         if tasks:
-            await asyncio.gather(*tasks)
+            # return_exceptions=True so one failing event handler doesn't abort
+            # the whole gather and silently drop the SIBLING events in the same
+            # sync response (a bare gather re-raises the first exception, leaving
+            # the rest of the batch unprocessed). Mirrors the invite/redaction
+            # gathers above. Surface each failure instead of swallowing it.
+            results = await asyncio.gather(*tasks, return_exceptions=True)
+            for result in results:
+                if isinstance(result, Exception):
+                    logger.warning(
+                        "Matrix: event handler failed during sync dispatch: %s",
+                        result,
+                    )
 
     def _is_self_sender(self, sender: str) -> bool:
         """Return True if the sender refers to the bot's own account.
diff --git a/tests/gateway/test_matrix.py b/tests/gateway/test_matrix.py
index 748422045..d239728b7 100644
--- a/tests/gateway/test_matrix.py
+++ b/tests/gateway/test_matrix.py
@@ -5276,3 +5276,36 @@ class TestDeviceIdRecoveryOnReconnect:
         assert None not in _verify_call.args[0]["@bot:example.org"]
 
         await adapter.disconnect()
+
+
+class TestMatrixDispatchSyncIsolation:
+    """A failing mautrix event handler must not abort the whole sync batch.
+
+    ``_dispatch_sync`` gathers the per-event handler tasks. Without
+    ``return_exceptions=True`` the first exception aborts the gather and the
+    sibling events in the same sync response are silently dropped.
+    """
+
+    @pytest.mark.asyncio
+    async def test_dispatch_sync_isolates_failing_handler(self, caplog):
+        import logging
+
+        adapter = _make_adapter()
+        ran = {"ok": False}
+
+        async def _boom():
+            raise RuntimeError("handler boom")
+
+        async def _ok():
+            ran["ok"] = True
+
+        client = MagicMock()
+        client.handle_sync = MagicMock(return_value=[_boom(), _ok()])
+        adapter._client = client
+
+        with caplog.at_level(logging.WARNING):
+            # Must not raise despite the failing handler.
+            await adapter._dispatch_sync({"next_batch": "s1"})
+
+        assert ran["ok"] is True  # the sibling handler still ran
+        assert "event handler failed" in caplog.text  # failure surfaced, not swallowed