The remarkable aspect of mainstream economists is their persistence treating values in (digital) ledgers as actual instances of banknotes (physical paper) as their mental model, then selectively dropping the model.
These are really good questions that actually would be best addressed by agent-based modeling, where individual agents (banks, employees, investors, bond traders, federal govt, state govt, firms etc) each have behaviors defined by rules and you run the system adjusting the rules to see how the scenario plays out. #julialang is a great tool for this using the Agents.jl library in part because it's fast