Comment 2 for bug 1359018

I thought maybe under the hood eventlet's green HTTPConnection or httplib's base were calling connect in __init__; but no...

They've all got this autoconnect thing going on where it'll make the connection when you first try to send data. Which for REPLICATE is basically once you've the whole request ready to go.

That is to say, I don't think we can do much to catch a connection timeout before we try to send the request. We could try to make some expect-100 something work on the REPLICATE verb; but since it's always been this way I think the smartest thing may just be to remove the useless ConnectionTimeout and the one silly `test_repl_to_node_http_connect_fails` that seemed to think a socket was getting opened in there and it would somehow magically return None on failure.