gistfile1.txt

public ActionResult Index() {

  return ProgressFlushResult(x => {
      x.PartialView("First");
      x.PartialView("Second");
      x.PartialView("Third");
  });

}

You've already done all the "work" (DB queries or service calls) by the time you hit return.

I'm thinking of moving towards:

[FlushEarly("Head")] // This could be a controller or global level filter
public ActionResult Index() {
    // do work

    return PartialView("Remainder");
});

The [FlushEarly("Head")] would run before the first line of code in the action result.

Of course, the Flush(ActionResult) method I proposed in my blog would still be required when you were using a dynamic head.

But for my money, yield return is still the most elegant way to do this.

You essentially want to create a "Workflow", which at the end of each state/node you want to flush. The yield return might be confusing to devs who might be familiar with that construct as it pertains to an "enumerable" thing.

I personally would prefer a more explicit implementation to something like this so it sticks out.