[tor-relays] Re: the case for relay state persistency