If I'm not mis-remembering, there's an interesting section in Dawkins' The Selfish Gene about cooperative grooming behavior (I think it was in some sort of water fowl) and how the birds deal with cheating behavior (individuals that accept grooming without reciprocating). My takeaway was roughly:

Any exchange of value based on trust is exploitable. The simplest cure is excluding the exploiter, but this doesn't scale well. The exploiter can skate on anonymity if the community is large enough to continually prey on someone new. Spreading news of an exploiter's behavior to others can greatly improve how well this scales, but this behavior also requires trust.

I think the more direct problem is with scale, and that the internet is at the nexus of many trust issues only because it has ramped up the scale and scope of many interactions.

I'm not super optimistic on solving this intrinsic problem of trust in social exchanges, but I do see this framing as a silver lining. It seems at least plausible to iterate offline at significantly smaller scales on mechanisms for building and maintaining trust--and rectifying its breaches--in ways that do actually scale.