Anyone else seeing 40% packetloss to 50BTC at the moment?
Also seems to be load issues at every new block on the network - my shares queue for anything up to a minute, and if there's a couple of blocks in succession I can lose a good few minutes of hashing because they all end up rejected. It leads to a very visible dip in my worker hashrate graph

Are you using the getwork/longpolling port 8332? I had this same problem but switched to stratum (3333) and have subsequently had no issues.