During a disk LUN failover, under certain conditions, virtual machines might stop responding for approximately one minute.
This issue is found in ESX 3.x and 4.x versions.
This issue is design specific. In x86 and x64 architectures, disk subsystems are prone to failures and allows transient failures and retrying, but the architecture does not allow retrying transient failures in the memory subsystem.
When transient failures of the underlying disk occur, the guest is not informed of these failures. Instead, by design, the guest operations are postponed until
the disk subsystem recovers completely.
This issue is design specific.