What do we know about actual limits of wire heading in humans?

Reward hacking, or wireheading, is a recurring theme in the Reinforcement Learning and Artificial General Intelligence literature, where an agent learn to exploit the system to maximise reward.

What do we know about actual limits of wire heading in humans? Do the body has an intrinsic defense mechanism to stop it?

For instance (wild guess) the intense electric signals or other side effects of wireheading could generate pain, up to a threshold in which the subject prefer to stop rather than keep hacking.

References of actual experiments are welcome.