Error 2002 is a server side issue. Proof provided

[Neon21]

Yes. In my experience, the error is mostly on the server side especially when the queue is large. They only provisioned 21k queue spots total for 8 worlds on each datacenter. But on the prime time, I frequently see the very large queue size (e.g., 4k) just for my world, which means that 21k queue spots are clearly not sufficient for 8 worlds. And that is usually the time where error 2002 is the most frequent.

From the linked doc, it is apparent that clients regularly tries to reconnect with the server. If the server has more than 21k connections at that point, we are likely to get 2002 errors. To make matters even worse, it is highly likely that I get 2002 error once again even if I restart my client to reconnect. If I end up getting 2002 errors several times in a row, I lose my spot in the queue and have to start from the beginning.

Three things need to be fixed:
(1) Increase the time window for reclaming a queue spot
(2) Even if the client fails to reconnect the server, let it automatically retry for the connection at a regular interval, rather than simply emitting errors and exiting the client.
(3) Further increase the login server queue size, at least for popular data centers. 21k is clearly not enough, since it only leaves about 2.6k queue spots for each world.