so what would you suggest, other than the following (I am using Phoenix 1.50):
${HOMEDIR}/phoenix-1.50/phoenix.py -u http://${MINERUSER}:${MINERPASS}@${$MINERSERV} -q 3 -k phatk VECTORS BFI_INT AGGRESSION=12 WORKSIZE=128 DEVICE=${1}
where $MINERSERV is either uswest or uscentral.
Both miners on Xubuntu boxen, one is catalyst 11.5, the other is 11.6; both use APP SDK 2.4. (that shouldn't affect the behavior that is occurring today with BTCGuild)
It seems that the part you are interested in is the
-q (queue) size. You setted it to 3 now ... try to increase it but keep in mind that the pool server will check if the result the miner submitted is actually a result of a work requested by that miner. I think is better to ask the pool owner how many requests the server records per miner. (If the server checks only at the last getwork request the -q configuration is useless).
BTW, another interesting thing is that weird network behavior with the mining server correlates well with freezes of the card (it is a 5830). IE long periods at Zero increase the likelihood that the card will freeze the OS.
Try to use some software to monitor the health of your card and some software to monitor the http traffic miners do.