Quote Originally Posted by Brennus View Post
Ah thank you. Very interesting and illuminating Wireshark analysis. The evidence seems pretty overwhelming, that our client software is at fault, not the login server at SE's end. Ergo this could easily be patched, presumably lessening the end-user frustrations from 2002 errors and losing our spots in the queue.
The problem unfortunately exists on both ends.

Login server capacity is limited to a finite number of connections. When your connection attempts to reestablish client-side as noted in the doc, it creates a race condition wherein it's highly likely that someone else has already taken the slot you just created while reconnecting, which results in you then being unable to connect to the login server which is now at capacity. Every 15 minutes you essentially have to pray that there is a 'slot' available in the login server as well.