That's what the earlier issue was at BTCGuild.... miners with very low latency had problems...
According to Ckolivas it had to do with the Java backend of PoolservJ.... the only solution he suggested was to include --net-delay, but that didn't help any
I have no clue though how much the 'íntroduced' delay is with --net-delay
Weird thing is it runs so well for a few hours but the longer the pool is up, the bigger the issues get.... for some users....
As I said before... I don't like it, but even with 5% stales I'm still getting way more than at most pools.... and then there's always the upside of lower rejects
