In my Windows 8.1 LAN consisting of 14 computers, there are several servers which are dedicated to very specific tasks. (For example is the "Save Server" responsible to run a daily batch script, which does connect to various clients and robocopies relevant files from there to its Storage Space.)
To start such a specialized machine, all that is required is that the operator remotely starts the machine in question (by pressing a command button in a native application, which provides power via a native home automation system). Such, the selected machine receives power, starts up as per UEFI settings, and auto-executes its script (which on some of the machines, for example on the "Save Server", also contains a Shutdown command when all is done). Typically, the on-time of these machines is 15 to 120 minutes a day.
In order to run a batch script on power-on, these servers have no password for the default user. (If there was one, a person would need to appear physically in front of the machine and enter the credentials, which is unfeasible.)
If it is "guaranteed" that there is no internal threat, is the omission of a password a vulnerability, exploitable from the outside?
After all, the "outside world" first needs to get through my Internet gateway which is (strongly) password-protected.
(Or put in other words, how do Google et al. handle their server farms? I can not imagine, that all those machines are password-protected, individually.)