diff --git a/spacy/language.py b/spacy/language.py
index 869fa09a7..16aa4967e 100644
--- a/spacy/language.py
+++ b/spacy/language.py
@@ -15,6 +15,7 @@ import multiprocessing as mp
 from itertools import chain, cycle
 
 from .tokenizer import Tokenizer
+from .tokens.underscore import Underscore
 from .vocab import Vocab
 from .lemmatizer import Lemmatizer
 from .lookups import Lookups
@@ -853,7 +854,10 @@ class Language(object):
         sender.send()
 
         procs = [
-            mp.Process(target=_apply_pipes, args=(self.make_doc, pipes, rch, sch))
+            mp.Process(
+                target=_apply_pipes,
+                args=(self.make_doc, pipes, rch, sch, Underscore.get_state()),
+            )
             for rch, sch in zip(texts_q, bytedocs_send_ch)
         ]
         for proc in procs:
@@ -1108,16 +1112,18 @@ def _pipe(docs, proc, kwargs):
         yield doc
 
 
-def _apply_pipes(make_doc, pipes, reciever, sender):
+def _apply_pipes(make_doc, pipes, receiver, sender, underscore_state):
     """Worker for Language.pipe
 
     receiver (multiprocessing.Connection): Pipe to receive text. Usually
         created by `multiprocessing.Pipe()`
     sender (multiprocessing.Connection): Pipe to send doc. Usually created by
         `multiprocessing.Pipe()`
+    underscore_state (tuple): The data in the Underscore class of the parent
     """
+    Underscore.load_state(underscore_state)
     while True:
-        texts = reciever.get()
+        texts = receiver.get()
         docs = (make_doc(text) for text in texts)
         for pipe in pipes:
             docs = pipe(docs)
diff --git a/spacy/tests/doc/test_underscore.py b/spacy/tests/doc/test_underscore.py
index 2877bfeea..c1eff2c20 100644
--- a/spacy/tests/doc/test_underscore.py
+++ b/spacy/tests/doc/test_underscore.py
@@ -7,6 +7,15 @@ from spacy.tokens import Doc, Span, Token
 from spacy.tokens.underscore import Underscore
 
 
+@pytest.fixture(scope="function", autouse=True)
+def clean_underscore():
+    # reset the Underscore object after the test, to avoid having state copied across tests
+    yield
+    Underscore.doc_extensions = {}
+    Underscore.span_extensions = {}
+    Underscore.token_extensions = {}
+
+
 def test_create_doc_underscore():
     doc = Mock()
     doc.doc = doc
diff --git a/spacy/tests/matcher/test_matcher_api.py b/spacy/tests/matcher/test_matcher_api.py
index e4584d03a..a826a0a0e 100644
--- a/spacy/tests/matcher/test_matcher_api.py
+++ b/spacy/tests/matcher/test_matcher_api.py
@@ -6,6 +6,7 @@ import re
 from mock import Mock
 from spacy.matcher import Matcher, DependencyMatcher
 from spacy.tokens import Doc, Token
+from ..doc.test_underscore import clean_underscore
 
 
 @pytest.fixture
@@ -200,6 +201,7 @@ def test_matcher_any_token_operator(en_vocab):
     assert matches[2] == "test hello world"
 
 
+@pytest.mark.usefixtures("clean_underscore")
 def test_matcher_extension_attribute(en_vocab):
     matcher = Matcher(en_vocab)
     get_is_fruit = lambda token: token.text in ("apple", "banana")
diff --git a/spacy/tests/regression/test_issue4849.py b/spacy/tests/regression/test_issue4849.py
index 834219773..85d03fe9a 100644
--- a/spacy/tests/regression/test_issue4849.py
+++ b/spacy/tests/regression/test_issue4849.py
@@ -3,6 +3,7 @@ from __future__ import unicode_literals
 
 from spacy.lang.en import English
 from spacy.pipeline import EntityRuler
+from spacy.tokens.underscore import Underscore
 
 
 def test_issue4849():
diff --git a/spacy/tests/regression/test_issue4903.py b/spacy/tests/regression/test_issue4903.py
new file mode 100644
index 000000000..9a3c10d61
--- /dev/null
+++ b/spacy/tests/regression/test_issue4903.py
@@ -0,0 +1,45 @@
+# coding: utf8
+from __future__ import unicode_literals
+
+import spacy
+from spacy.lang.en import English
+from spacy.tokens import Span, Doc
+from spacy.tokens.underscore import Underscore
+
+
+class CustomPipe:
+    name = "my_pipe"
+
+    def __init__(self):
+        Span.set_extension("my_ext", getter=self._get_my_ext)
+        Doc.set_extension("my_ext", default=None)
+
+    def __call__(self, doc):
+        gathered_ext = []
+        for sent in doc.sents:
+            sent_ext = self._get_my_ext(sent)
+            sent._.set("my_ext", sent_ext)
+            gathered_ext.append(sent_ext)
+
+        doc._.set("my_ext", "\n".join(gathered_ext))
+
+        return doc
+
+    @staticmethod
+    def _get_my_ext(span):
+        return str(span.end)
+
+
+def test_issue4903():
+    # ensures that this runs correctly and doesn't hang or crash on Windows / macOS
+
+    nlp = English()
+    custom_component = CustomPipe()
+    nlp.add_pipe(nlp.create_pipe("sentencizer"))
+    nlp.add_pipe(custom_component, after="sentencizer")
+
+    text = ["I like bananas.", "Do you like them?", "No, I prefer wasabi."]
+    docs = list(nlp.pipe(text, n_process=2))
+    assert docs[0].text == "I like bananas."
+    assert docs[1].text == "Do you like them?"
+    assert docs[2].text == "No, I prefer wasabi."
diff --git a/spacy/tests/regression/test_issue4924.py b/spacy/tests/regression/test_issue4924.py
index 8aea2c3d5..0e45291a9 100644
--- a/spacy/tests/regression/test_issue4924.py
+++ b/spacy/tests/regression/test_issue4924.py
@@ -11,6 +11,6 @@ def nlp():
     return spacy.blank("en")
 
 
-def test_evaluate(nlp):
+def test_issue4924(nlp):
     docs_golds = [("", {})]
     nlp.evaluate(docs_golds)
diff --git a/spacy/tokens/underscore.py b/spacy/tokens/underscore.py
index b36fe9294..8dac8526e 100644
--- a/spacy/tokens/underscore.py
+++ b/spacy/tokens/underscore.py
@@ -79,6 +79,14 @@ class Underscore(object):
     def _get_key(self, name):
         return ("._.", name, self._start, self._end)
 
+    @classmethod
+    def get_state(cls):
+        return cls.token_extensions, cls.span_extensions, cls.doc_extensions
+
+    @classmethod
+    def load_state(cls, state):
+        cls.token_extensions, cls.span_extensions, cls.doc_extensions = state
+
 
 def get_ext_args(**kwargs):
     """Validate and convert arguments. Reused in Doc, Token and Span."""