I don't mind the model for places where the partisan outcome is more or less a foregone conclusion. It would make the general an actual contest between two more evenly matched contestants, at least in theory. But I've got a problem putting most of the context back to a time (whether in spring or in summer, primaries have lower turnout than general elections) when fewer people are engaged with the process.
Of course the vast majority of districts are a foregone conclusion in terms of the parties. We tend to focus on the handful of true swing seats, but they are not the norm.
Sure, I know. But that's where the action is, and designing a system that makes those races easier to game at the primary stage doesn't help anything.