Connections to NFS are failing, cannot ping the local gateway

Issue

Randomly various members of a cluster loose connections to the NFS server. We have noticed that pings to the local gateway needed to reach the NFS server also fail. Connection is via a mode 4 (802.3ad) bond with layer3+4 hashing. We have further observed that tcpdump shows that the ping requests leave the system on slave-1 but that no ping requests are observed on traces collected on the network. Disabling and re-enabling slave-1 resolves the problem.