[18:59:50] <AaronSchulz>	 Krinkle: I probably just make the GET_LOCK() also get the timestamp since all those _CRITICAL methods do that anyway
[19:00:05] <AaronSchulz>	 then there is no extra rountrip
[19:03:22] <Krinkle>	 that aviods the latency but presumably not the complexity. if someone else submits a patch in a month's time removing all that in favour of something simple, we should be able to explain why we can't accept that.