back then, i tried using urllib.request (from the standard library, not urllib3) but it lacks what Requests/urllib3 has -- connection pools and keep alive [or that's where i thought the magic was] -- so my code ran much slower.

it turns out that urllib.request uses http.client, but it closes the connection. so by using http.client directly, i can keep the connection open [that's where the real magic is]. now my code runs as fast as Requests/urllib3, but in 5 lines of code instead of 4,000-15,000+

moral of the story: RTFM over and over and over again.

  """Fetch users from the Okta API and paginate."""
  
  import http.client
  import json
  import re
  import urllib.parse
  
  # Set these:
  host = 'domain.okta.com'
  token = 'xxx'
  url = '/api/v1/users?' + urllib.parse.urlencode({'filter': 'profile.lastName eq "Doe"'})
  
  headers = {'authorization': 'SSWS ' + token}
  conn = http.client.HTTPSConnection(host)
  while url:
     conn.request('GET', url, headers=headers)
     res = conn.getresponse()
     for user in json.load(res):
         print(user['id'])
     links = [link for link in res.headers.get_all('link') if 'rel="next"' in link]
     url = re.search('<https://[^/]+(.+)>', links[0]).group(1) if links else None

https://docs.python.org/3/library/http.client.html