I don't have the patience to work through most papers on causal deviance. This isn't an attempt to solve any of the problems having to do with deviant causal chains, only an attempt to say something about the significance of deviant causal chains.
The problem that deviant causal chains pose for a Davidsonian approach to action is one of the motivations for Steward's agency incompatibilism. As she sees it, actions just aren't the sorts of things that have causal antecedents. It's a mistake to think of actions as events that have been brought about in the right way by the agent's states of mind or mental events.
Here's a worry. It seems that you'll run into problems having to do with deviant causal chains in trying to give a causal account of believing for a reason, but it doesn't seem like beliefs are the sorts of things that shouldn't have causal antecedents. Agency incompatibilism has its merits, but I don't think that determinism is a threat to belief.
On Marcus' account, it looks like acting for a reason and believing for a reason are given the same treatment--to believe for a reason or act for a reason is to represent the belief or action as to be believed or done in light of something else. On one way of reading this, deviant chains cannot arise because that requires a causal chain between two distinct existences. A virtue of Marcus' approach is that it does away with the distinct existences. Believing for a reason doesn't involve a causal relation between a reason or some reasons and a belief. It involves representing the fact/proposition believed as to be believed in light of something else and that representation isn't a relation between two distinct existences.
Here are two worries. First, Marcus' account cannot be applied to the case of non-inferential belief. Thus, the account seems insufficiently general. Whatever story we tell about believing for a reason in the non-inferential case, we'll need to rule out deviancy in this case without using his trick. You might think that whatever we do to understand the relationship between the reason for which you believe and your belief in, say, the case of perceptual knowledge, we'll be able to tell a similar story in the case of inferential knowledge. Second, there are deviant causal chains that don't have to do with belief, action, or reasons at all. There are cases of that show that the connection between, say, the fragile glass' shattering and the striking of the glass isn't right for the glass to have shattered because of its fragility. (I think I owe this point to Hyman).