diff --git a/conftest.py b/conftest.py
index 5737dae00..ba169cf87 100644
--- a/conftest.py
+++ b/conftest.py
@@ -1,3 +1,4 @@
+import sys
 from pathlib import Path
 
 import pytest
@@ -8,8 +9,6 @@ def _py_files(folder):
 
 
 collect_ignore = [
-    # no tests and not importable on 3.5
-    "scrapy/utils/asyncgen.py",
     # not a test, but looks like a test
     "scrapy/utils/testsite.py",
     # contains scripts to be run by tests/test_crawler.py::CrawlerProcessSubprocess
@@ -20,6 +19,10 @@ collect_ignore = [
     *_py_files("tests/py36"),
 ]
 
+if sys.version_info < (3, 6):
+    # not importable on 3.5
+    collect_ignore.append("scrapy/utils/asyncgen.py")
+
 for line in open('tests/ignores.txt'):
     file_path = line.strip()
     if file_path and file_path[0] != '#':
diff --git a/docs/topics/coroutines.rst b/docs/topics/coroutines.rst
index d4f9af86b..3cf005e28 100644
--- a/docs/topics/coroutines.rst
+++ b/docs/topics/coroutines.rst
@@ -44,7 +44,8 @@ hence use coroutine syntax (e.g. ``await``, ``async for``, ``async with``):
 
 -   The
     :meth:`~scrapy.spidermiddlewares.SpiderMiddleware.process_start_requests`
-    method of :ref:`spider middlewares <custom-spider-middleware>`.
+    method of :ref:`spider middlewares <custom-spider-middleware>`. See
+    :ref:`async-start_requests` below for details.
 
 -   :ref:`Signal handlers that support deferreds <signal-deferred>`.
 
@@ -57,11 +58,17 @@ hence use coroutine syntax (e.g. ``await``, ``async for``, ``async with``):
 Asynchronous start_requests and spider middlewares
 ==================================================
 
-.. versionadded:: 2.1
+.. versionadded:: 2.2
 
 The :meth:`~scrapy.spiders.Spider.start_requests` spider method can be an
-asynchronous generator. In this case all spider middlewares used with this
-spider that have the
+asynchronous generator::
+
+    async def start_requests():
+        # ...
+        yield scrapy.Request(...)
+        # ...
+
+In this case all spider middlewares used with this spider that have the
 :meth:`~scrapy.spidermiddlewares.SpiderMiddleware.process_start_requests`
 method must support this: if they receive an asynchronous iterable, they must
 return one as well. On the other hand, if they receive a normal iterable, they
@@ -83,15 +90,15 @@ Here is an example of a universal middleware using this approach::
 
     class ProcessStartRequestsAsyncGenMiddleware:
         async def process_start_requests(self, start_requests, spider):
-            async for r in as_async_generator(start_requests):
-                # ... do something with r
-                yield r
+            async for req in as_async_generator(start_requests):
+                # ... do something with req
+                yield req
 
 If this method includes asynchronous code, that code will work even with
 synchronous :meth:`~scrapy.spiders.Spider.start_requests`.
 
 Another option is to make separate methods for normal and asynchronous
-iterables and choose one at the run time::
+iterables and choose one at run time::
 
     class ProcessStartRequestsAsyncGenMiddleware:
         def _normal_process_start_requests(self, start_requests, spider):
diff --git a/docs/topics/spider-middleware.rst b/docs/topics/spider-middleware.rst
index 0cf3e1ea8..0e214a40a 100644
--- a/docs/topics/spider-middleware.rst
+++ b/docs/topics/spider-middleware.rst
@@ -148,8 +148,8 @@ object gives you access, for example, to the :ref:`settings <topics-settings>`.
     .. method:: process_start_requests(start_requests, spider)
 
         .. versionadded:: 0.15
-        .. versionchanged:: 2.1
-           Since 2.1 this can take and return an :term:`python:asynchronous
+        .. versionchanged:: 2.2
+           Since 2.2 this can take and return an :term:`python:asynchronous
            iterable`.
 
         This method is called with the start requests of the spider, and works
@@ -165,7 +165,7 @@ object gives you access, for example, to the :ref:`settings <topics-settings>`.
         middleware method should be an :term:`python:asynchronous generator`,
         which should support both synchronous and asynchronous
         ``start_requests``. :func:`scrapy.utils.asyncgen.as_async_generator`
-        can be used to convert both kinds of ``start_requests`` to an
+        can be used to convert both kinds of ``start_requests`` into an
         asynchronous iterable.
 
         .. note:: When implementing this method in your spider middleware, you