The solution here seems to generally work:

• https://stackoverflow.com/questions/22492484/how-do-i-get-the-ip-address-from-a-http-request-using-the-requests-library

# Basically works, except the edge case shown below
r.server_ip           = r.raw._connection.sock.getpeername()[0]     
r.server_port         = r.raw._connection.sock.getpeername()[1]    

However, there seems to be some edge case where this occassionally doesn't work. See these debugging statements:

Example of Failure

dict_keys(['reason', 'request', 'status_code', 'raw', '_content_consumed', 'url', 'encoding', 'connection', 'history', 'elapsed', 'headers', '_content', 'cookies'])
r.status_code              200
r.raw                      <requests.packages.urllib3.response.HTTPResponse object at 0x7fd9f4069d68>
r.raw._connection          <requests.packages.urllib3.connection.HTTPConnection object at 0x7fd9e4783ac8>
r.raw._connection.sock     None
r.history                  []
r.elapsed                  0:00:00.937903
r.request                  <PreparedRequest [GET]>
r.connection               <requests.adapters.HTTPAdapter object at 0x7fd9e47e5f98>
r.headers                  {'Content-Encoding': 'br', 'Content-Disposition': 'attachment; filename="file.txt"', 'Date': 'Wed, 25 Sep 2019 22:21:36 GMT', 'X-Cache': 'MISS from barracuda.greenetwp.us', 'Pragma': 'no-cache', 'Via': '1.0 barracuda.greenetwp.us:8080 (http_scan/4.0.2.6.19)', 'Cache-Control': 'no-cache, must-revalidate', 'X-Frame-Options': 'SAMEORIGIN', 'Expires': '-1', 'Content-Type': 'text/javascript; charset=UTF-8', 'X-XSS-Protection': '0', 'Server': 'gws', 'Proxy-Connection': 'close'}
r.reason                   OK
r.cookies                  <RequestsCookieJar[]>
r._content_consumed        False
r._content                 False

Example of Success

dict_keys(['server_ip', 'reason', 'server_port', 'request', 'status_code', 'raw', '_content_consumed', 'url', 'encoding', 'connection', 'history', 'elapsed', 'headers', '_content', 'cookies'])
r.status_code              200
r.raw                      <requests.packages.urllib3.response.HTTPResponse object at 0x7fd9e47e51d0>
r.raw._connection          <requests.packages.urllib3.connection.HTTPConnection object at 0x7fd9e47a0fd0>
r.raw._connection.sock     <socket.socket fd=63, family=AddressFamily.AF_INET, type=2049, proto=6, laddr=('192.168.1.154', 47000), raddr=('36.55.230.146', 8888)>
r.history                  []
r.elapsed                  0:00:00.592083
r.request                  <PreparedRequest [GET]>
r.connection               <requests.adapters.HTTPAdapter object at 0x7fd9e47a0ac8>
r.headers                  {'Content-Encoding': 'gzip', 'Content-Disposition': 'attachment; filename="file.txt"', 'Date': 'Wed, 25 Sep 2019 22:21:36 GMT', 'Pragma': 'no-cache', 'Cache-Control': 'no-cache, must-revalidate', 'Server': 'nginx/1.14.0', 'X-Frame-Options': 'SAMEORIGIN', 'Connection': 'keep-alive', 'Expires': '-1', 'Content-Type': 'text/javascript; charset=UTF-8', 'X-XSS-Protection': '0', 'Transfer-Encoding': 'chunked'}
r.reason                   OK
r.cookies                  <RequestsCookieJar[]>
r._content_consumed        False
r._content                 False

The only obvious difference between the requests that succeeded and the ones that failed was that the failed requests had a Content-Encoding of 'br' (Brotli). It's possible this is somehow connected, since there are known issues related to support for Brotli in the requests library:

• https://github.com/request/request/issues/2907
• https://github.com/request/request/issues/2866

Anyway, using this version instead of the code at the start of this post seems to work consistently:

• https://myshell.co.uk/blog/2017/06/python-get-remote-ip-from-http-request-using-the-requests-module/

# Seems to work consistently 
r.server_ip           = r.raw._original_response.fp.raw._sock.getpeername()[0]
r.server_port         = r.raw._original_response.fp.raw._sock.getpeername()[1]

Sir, I can do you a nice SEO.