SRP + ISP used to their conclusion gives one-method interfaces and very very simple classes --> this seems a bit verbose for its intent
(Pure) Function: that's a behavior without data (everything needed for the computation is passed in the function arguments
Closure = Function + environment --> this is behavior with data...
Closure / Object is an example of Duality
The author's conclusion is to use F# when attaining the limits of SRP/ISP...
I'd rather stick to C# but use loose forms of class extensibility such a Extension methods (traits, mixins...): they provide a container for functional-style code and are bound to public (well... internal as well) APIs only.