diff --git a/engineio/payload.py b/engineio/payload.py
index fe15d76d..25bd678e 100644
--- a/engineio/payload.py
+++ b/engineio/payload.py
@@ -33,7 +33,6 @@ def encode(self, b64=False):
 
     def decode(self, encoded_payload):
         """Decode a transmitted payload."""
-        fixed_double_encode = False
         self.packets = []
         while encoded_payload:
             if six.byte2int(encoded_payload[0:1]) <= 1:
@@ -49,36 +48,20 @@ def decode(self, encoded_payload):
                 i = encoded_payload.find(b':')
                 if i == -1:
                     raise ValueError('invalid payload')
-                # the packet_len below is given in utf-8 characters, but we
-                # receive the payload as bytes, so down below this length is
-                # adjusted to reflect byte length
-                packet_len = int(encoded_payload[0:i])
-                if not fixed_double_encode:
-                    # the engine.io javascript client sends text payloads with
-                    # a double UTF-8 encoding. Here we try to fix that mess and
-                    # restore the original packet
-                    try:
-                        # first we remove one UTF-8 encoding layer
-                        fixed_payload = encoded_payload.decode(
-                            'utf-8').encode('raw_unicode_escape')
-
-                        # then we make sure the result can be decoded a second
-                        # time (this will raise an exception if not)
-                        fixed_payload.decode('utf-8')
 
-                        # if a second utf-8 decode worked, then this appears to
-                        # be a double encoded packet, so here we keep the
-                        # packet after a single decode, since the packet class
-                        # will perform a decode as well, and in this case it is
-                        # not necessary to adjust the packet length
-                        encoded_payload = fixed_payload
-                    except:
-                        # if we couldn't apply a double utf-8 decode then
-                        # the packet must have been correct, so we just adjust
-                        # the packet length to be in bytes and not utf-8
-                        # characters and keep going
-                        packet_len += len(encoded_payload) - len(fixed_payload)
-                    fixed_double_encode = True
-                pkt = encoded_payload[i + 1: i + 1 + packet_len]
+                # extracting the packet out of the payload is extremely
+                # inefficient, because the payload needs to be treated as
+                # binary, but the non-binary packets have to be parsed as
+                # unicode. Luckily this complication only applies to long
+                # polling, as the websocket transport sends packets
+                # individually wrapped.
+                packet_len = int(encoded_payload[0:i])
+                pkt = encoded_payload.decode('utf-8', errors='ignore')[
+                    i + 1: i + 1 + packet_len].encode('utf-8')
                 self.packets.append(packet.Packet(encoded_packet=pkt))
+
+                # the engine.io protocol sends the packet length in
+                # utf-8 characters, but we need it in bytes to be able to
+                # jump to the next packet in the payload
+                packet_len = len(pkt)
             encoded_payload = encoded_payload[i + 1 + packet_len:]
diff --git a/tests/test_payload.py b/tests/test_payload.py
index 642feb60..dbb90d24 100644
--- a/tests/test_payload.py
+++ b/tests/test_payload.py
@@ -50,18 +50,8 @@ def test_decode_invalid_payload(self):
         self.assertRaises(ValueError, payload.Payload,
                           encoded_payload=b'bad payload')
 
-    def test_decode_double_encoded_utf8_payload(self):
-        p = payload.Payload(encoded_payload=b'3:4\xc3\x83\xc2\xa9')
-        self.assertEqual(len(p.packets), 1)
-        self.assertEqual(p.packets[0].data.encode('utf-8'), b'\xc3\xa9')
-
-    def test_decode_double_encoded_utf8_multi_payload(self):
-        p = payload.Payload(encoded_payload=b'3:4\xc3\x83\xc2\xa94:4abc')
+    def test_decode_multi_payload(self):
+        p = payload.Payload(encoded_payload=b'4:4abc\x00\x04\xff4def')
         self.assertEqual(len(p.packets), 2)
-        self.assertEqual(p.packets[0].data.encode('utf-8'), b'\xc3\xa9')
-        self.assertEqual(p.packets[1].data, 'abc')
-
-    def test_decode_single_encoded_utf8_payload(self):
-        p = payload.Payload(encoded_payload=b'3:4\xc3\xa9')
-        self.assertEqual(len(p.packets), 1)
-        self.assertEqual(p.packets[0].data.encode('utf-8'), b'\xc3\xa9')
+        self.assertEqual(p.packets[0].data, 'abc')
+        self.assertEqual(p.packets[1].data, 'def')
diff --git a/tox.ini b/tox.ini
index b784257a..d4123e6f 100644
--- a/tox.ini
+++ b/tox.ini
@@ -38,7 +38,7 @@ deps=
 deps=
     flake8
 commands=
-    flake8 --exclude=".*" --ignore=E402 engineio tests
+    flake8 --exclude=".*" --ignore=E402,E722 engineio tests
 
 [testenv:docs]
 changedir=docs