GCP provides two options for floating IP addresses in Computer Engine: https://cloud.google.com/architecture/patterns-for-floating-ip-addresses-in-compute-engine#heartbeat-exposed and https://cloud.google.com/architecture/patterns-for-floating-ip-addresses-in-compute-engine#using_a_heartbeat_mechanism_to_switch_a_routes_next_hop
From my understanding, load balancer approach with heartbeat exposed approach can have better failover time performance, but it come with an extra cost.
The route-based approach also uses heartbeat between 2 server vms. Route is free but it may lead to longer failover time. However, this approach can help shorter failover time compared to other route-based approach. Since according to GCP doc, "You can switch the route priority pattern and don't have to delete and replace VMs through autohealing for failover."
Does this heartbeat mechanism route-based approach still have a longer failover time than the load balancer approach? Does load balancer generally provides better availablity?