Moved to NGFW for better support.
I would recommend opening a support ticket for this issue. It is possible the switches in surrounding network have problems handling the CVI MAC (e.g. if it changes to other port), but if the cluster has worked fine before in loadbalancing configuration and the switches are the same, it's unlikely this is the cause. Another point regarding clustering and CVI, I would recommend using packet dispatch CVI mode, in case you now have something else like multicast MAC.
Other possible explanation is that node 2 was in some error state, but determining this would require analyzing the sginfos by support.