two years later....

and race servers are still as unstable as ever. do you not prioritize racing/hardcore on better/separate instance servers from softcore to avoid explicit overloading of one instance?

connect to gateway at 13ms. move to random area and get sent to singapore or some other bfe zone and shoot between 120 and 600ms. of course you will not necessarily know this during the grace period because until actions start getting processed the true RTT is unknown. once you lose grace you rip.

what is the point of choosing a gateway when you never stay on it in solo play? so retarded. you should never leave your gateway unless you join a shared instance. if the rtt to a given instance is more than twice as bad as the previous instance then that transfer should be reverted until a closer instance is available either due to spawning a new one or finding an existing one that satisfies the latency condition.

one should not have to monitor latency and play chicken to avoid dying. the first rule of lockstep is you roll back any time the client and server lose track of state and are unable to commit actions during a transaction.

when a client shoots to over 1800ms because the server instance goes unresponsive then the client should be force disconnected instead of ripping anon to standard league. there's no point to hardcore if a player can do nothing about crap instance infrastructure. your lockstep code doesn't allow for a client queue of actions to process when the server goes awol and play spool catchup afterwards. if it did then when i ran back to the start of the instance while waiting for the server to catch up i would have had a chance to survive. since i cant move until the server confirms each step it's actually worse to play under lockstep than it was with the previous predicate model until you have stable localized instances. stutter stepping is inane design.