I think we should keep in mind that:
  • The nub of the issue is the server live-state model SE chose for this game (see: OP), and that this model only performs well in a low-latency scenario, and is unplayable (game-breaking) otherwise.
  • Latency from most countries is simply too high, or packet loss is too important to get a good result (apparently more from/to the Canadian DataCenter here, compared to the JP DC). Consider that people living in Montreal still have issues dodging AoE's and can cast while running —with a ping of 20!
  • Placing more datacenters (EU, South Am., South-East Asia…) would make the issue less apparent but no less real for a number of players.

Also choosing TCP over UDP is a bad choice for games since TCP is used for non-time critical applications, whereas UDP is used for games and applications that require fast transmission of data (its stateless nature is useful for servers that receive large amount of small queries, which is exactly what MMORPG's are).

Therefore the only definitive solution would be to change the netcode architecture from server live-state to client live-state, and probably move to UDP, as 99% other MMO's (and fast-paced online games in general) do out there. The quality and reliability of internet connections in most countries is simply not up to the task of sustaining a server live-state model in 2013. Add to that the fact that many ISP's are confronted to global internet traffic saturation and resort to throttling bandwidth and request delay (TCP only? I'd suppose, but don't know), which further compounds the issue (which is probably why using VPN's may help some people, should their ISP be a lousy, but doesn't entirely solve the issue).