Great post and great thinking.
I wonder if you could really implement feeling of pain to robots, wouldn't it's decisions be more strict or black&white-ish? As it would make choices in order to avoid the unnecessary pain caused to themselves. Or perhaps they would consider it as a weakness and immediately by self-learning patterns will "upgrade" themselves in such a way that they will block these "feelings". Something comparable to what people were doing in the 2002 movie Equilibrium.
From a technological point of view, I read that it is indeed possible, in a certain way. You also make an interesting point that they could see this as weakness and upgrade themselves to not feel pain. Interesting!