I had a question this week about what would happen if disk.terminateVMonPDLDefault was enabled and das.maskCleanShutdownEnabled would be disabled. I wrote about many stretched cluster scenarios in the past and this customer wanted to know if there would be a reason to configure it that way.
The answer is short and sweet. if you enable disk.terminateVMonPDLDefault and don't enable das.maskCleanShutdownEnabled then your VMs will be killed when a PDL is issued. However as HA believes that the VM is cleanly shutdown it will not make an attempt to restart them.
Once again, please do note that disk.terminateVMonPDLDefault only acts when a PDL is issued. When your datastore has an "APD" (All Paths Down) condition your VMs will not be killed!
This white paper about vSphere Metro Storage Clusters that was released a couple of weeks back goes in to more depth around these failure scenarios.