IBM Clusters have a technique of even taking over the IP address of the failing system.
This allows inbound traffic to lose at most only one message and the retry will go
to the backup system.
Clustered systems run duplication programs; eg if systemA has programs{a1,b1,c1,...}
then the shadow programs on systemB are {a2,b2,c2,...} so that when the traffic
is redirected, the required services are already primed to receive that traffic.
The 'hot-backup' implementation causes traffic and processing in a round-robin
technique so that the loss of systemA is a non-issue.
The Small Business or Home user can not get to this level of duplication without
mortgaging the farm, but if you have a domain controller environment, some
smart code could force DHCP to reassign IP address.
IMO, that's about as far as PC clusters could go.