FT uses network heartbeats and some sort of "atomic operation" on VMFS to know if the VM failed or the host failed.
The reason HA is not FT resides on the concept. HA needs to be sure that won't cause a mess to power on VMs again in recovery. The right question is when FT will have so much functionality that will cause HA to be useless. Until now is very difficult to maintain an FT level on a VMware cluster (and, in fact, any fault tolerance solution I know almost need to duplicate resources), and we bet in a near future the replication of CPU, memory and disk have a better development, avoiding the resource waste caused by the actual ways to do the whole process.
Marcelo Soares
VMWare Certified Professional 310/410
Technical Support Engineer
Globant Argentina
Consider awarding points for "helpful" and/or "correct" answers.
Marcelo Soares