To address Andrew's comment number 3 (we refer to fig 1 which is a gARG), my thoughts (we don't want to emphasise gARG vs eARG too much here, as you could create a prospective eARG, even though people don't, for good reason), and (partially) Gideon's comments about when prospective ARGs might be useful (I just mentioned forward simulations higher up).
No need to merge this yet, as I know that Jerome may disagree with some of it, but PRed so that we have something on which to base a discussion.
To address Andrew's comment number 3 (we refer to fig 1 which is a gARG), my thoughts (we don't want to emphasise gARG vs eARG too much here, as you could create a prospective eARG, even though people don't, for good reason), and (partially) Gideon's comments about when prospective ARGs might be useful (I just mentioned forward simulations higher up).
No need to merge this yet, as I know that Jerome may disagree with some of it, but PRed so that we have something on which to base a discussion.