Problem with a medium-level development server that's sometimes out of action, is that it may respond to ping but be otherwise dead to the world. I've seen unreliable servers act like a LaBrea tarpit.
True, a server can have the network stack functional while the rest of the OS is out to lunch.
However PING is a cheap way to avoid a machine that really is unavailable such as development or VM which are regularly shut down to reduce cost or conserve resources.
So you may still get broken machines which respond to PING and very little else, but at least you've narrowed the field before attempting more expensive activities.
