We have resolved this issue on a couple of Clusters by setting the following in Registry on the nodes(After having updated service packs, driver, firmware and disabling TCP/IP offload)

1. Set TcpMaxDataRetransmissions to 30 (decimal);

http://technet2.microsoft.com/WindowsServer/en/library/7dac9001-3e55-4e9c-b0fa-52841ece2fdd1033.mspx

2. Set KeepAliveInterval to 25000 (decimal).

http://technet2.microsoft.com/WindowsServer/en/library/734570a2-06d6-450e-b765-ccfa7530af491033.mspx

It worked for us you mileage may.