The expression "defection risk" implies the one-shot prisoner dilemma context and neither such a situation is common in real life, nor normal people think in such categories (correctly, too).

You're reading too much into what RomeoStevens wrote - at no point did he explicitly mention the one-shot Prisonner's Dilemma.

A pretty common usage here is to use the Prisonner's Dilemma as a simplified model (think spherical cow on a frictionless plane, or perfect gas) of many morally-relevant situations.

This model is not what people explicitly think about (just like people don't explicitly think about social status when they are outraged or dismissive, or don't explicitly think about expected utility when deciding), but it may still be a good (simplified) model of what people think.

And RomeoStevens is referring to what people think, he's just using "defection risk" as a shorthand. If you ask normal people, they'll usually talk in terms of trust.

You may object that the model is not good enough, but you'll need a better argument than "it's not what people think" (nobody is claiming it is); do you similarly object to discussing people's choices in terms of expected utility and opportunity costs?