Let me take a stab at a response here; I will run out of time, and then try to post more later tonight or tomorrow.
I think Doug makes an excellent point that "what, for example, people get an MA in at SAIS is called International Relations, and so is theorizing about, say, epistemic communities and hegemony of domestic systems of government" -- even though these activities have, as far as I can tell, less in common with one another than (for instance) the playing of American football and the Monday-morning quarterbacking of a game after it's been played out. At least the Monday-morning quarterbacks have the actual practice of on-field football as a point of reference; no such tight linkage is necessarily the case in IR.
Let me elaborate a bit. For me, "scholarship" is basically a synonym for "producing systematic knowledge that is in some sense valid." [There's a lot packed into that phrase "in some sense," and by "a lot" here I mean "the entire field of the philosophy of science and probably epistemology in general." but I want to set that aside for the moment.] Hence, a scholarly approach to anything involves generating systematic, in-some-sense-valid knowledge about it. This is a qualitatively different matter than actually engaging in the thing in question. Scholarship about baseball is not playing baseball; scholarship about music is not the performance of music; likewise scholarship about world politics is not "doing" world politics.
I don't think I've said anything particularly controversial yet. But where it gets controversial is the relationship between scholarship and object. We have two ideal-typical positions on this: scholarship ought to improve practice, and scholarship can't possibly improve practice, at least not directly. Rob clearly prefers door #1; I prefer door #2. Rob's position is the classic Enlightenment hope for the sciences of society: place practice on a more rational basis, achieve better results, produce a world that looks more like the world we want to live in; I think that's both dangerous and a little naive -- dangerous because it puts a potential transcendental justification for coercion in the hands of would-be reformers (after all, if the experts told us that we can do this, and you disagree, then you're either stupid or obstinate, and in either way you're in the way so forcibly removing you starts to look like a good idea) and naive because it presumes that scholarly knowledge translates more or less simply to the actual world (and once again, if it doesn't, maybe we ought to use force to make the world look more like the model . . .).
I prefer option #2 -- scholarship can't possibly improve practice, at least not directly -- in part because people claiming to have Reason/God/Truth on their sides ("Jesus/Buddha/Muhammad/science likes my policy better!") have been responsible for most of the senseless death in human history, in part because systematic scholarly knowledge is by nature an abstraction (and sometimes a severe abstraction, in which the actual practice of anyone in particular disappears -- the sports analogy here would be to sabremetric analyses of baseball, and we've seen what happens when actual baseball teams try to directly implement strategies that look valid sabremetrically) and therefore not fit for any sort of direct translation into practice, and in part because scholarly knowledge is irreducibly perspectival and thus does not seem to me to be a good solid basis for decision-making (although it can certainly inform decision-making as one element among others).
A concrete example. Rob references the Counter-Insurgency Manual; I can see two scholarly things to do with that work: conduct a discourse analysis of how it implicitly and explicitly enframes the issues of personhood and rights and the like (my preference), and conduct research on the manual's perhaps implicit claims about the success of various tactics to see if those claims stand up (maybe what Rob would prefer). Both of these kinds of scholarship can inform decision-making, the first by highlighting the ethical issues involved in our enframing of actions in one way rather than another, the second by giving some sense of likely results of courses of action. But in neither case do they tell someone involved in counter-insurgency what to do, let alone instruct someone in how to engage in counter-insurgency operations! In neither case does scholarship -- systematic, in=some-sense-valid knowledge -- solve the fundamental practical and political question of what someone ought to do in a specific situation.
To tie this back to MAs and practitioners: if I want to go into the practice of world politics, I want to learn how to make policy decisions. If I am teaching someone who wants to go into the practice of world politics, I want to give them a sense of the irresolvable dilemmas that they are going to face, and help them to develop a critical disposition that can help them grapple with those dilemmas. None of this has anything whatsoever to do with the systematic results of my scholarly investigations into anything; it has to do with exercises designed to clarify value-commitments and their implications.
If, on the other hand, I am advising policymakers, I probably want to present my results but then realize that it is not my job to make the tough decisions surrounding their implementation (as Weber said, politics is the slow boring of hard boards) and leave that to the policymakers. But that's not teaching students, it's offering a scholarly input to a policymaking process that a scholar has to remain independent of lest she or he compromise her or his detachment and turn into a partisan for one or another group or party (and thus, by definition, no longer be engaged in doing scholarship).
And in neither case does this have anything to do with "certifying idealism," which was my original point.
The part I promised to post later reads like this:
Three other things that occurred to me.
1)I still don't see the intellectual point of a terminal MA because, contra Doug, I don't see those classes at operating at a level any higher than what one finds in undergrad. In fact, I teach my MA theory course like I taught my undergraduate theory classes: we read Hobbes, Locke, etc., and talk through their arguments and implications thereof. Precisely what I did and do for undergrads. So from my perspective a terminal MA in IR looks like more undergrad plus a few "policy" classes (talk about some issues, generally in a completely a-theoretical way) and some "methods" classes (basic stats -- which they probably had in undergrad already anyway -- and sometimes "risk analysis," which to a social scientist like myself just looks like bad research design and flawed data analysis). Yes, I get that this helps people get jobs. What I don't get is why it helps people get jobs, and what people think that a graduate of such a program can possibly do that a well-educated undergrad can't already do.
2) I still maintain that the most helpful thing any of us in the academy can do for any of our students is to temper their idealism. This is especially true for Americans, who tend to be pretty unreflective idealists by default. This is why political realism is actually a critical theory in the US context, since it maintains that there are limits on political possibilities, limits that can't be overcome with a little effort and a clever slogan. Again, I think that this is no different than a good undergraduate program, but if MA students are not getting that until their post-graduate work, I suppose better late than never.
3) I pity the student who comes to me hoping to be trained in job-relevant skills if their anticipated job is someplace other than academia. Academia is where I work, and I know how to do that job pretty well, so I can pass on bits of practical advice and professional wisdom. The State Department? I can find it on a map, but I've never worked there and have no desire to do so, so I am not likely to be of any use to students looking to be trained in how to succeed at State (or in any other DC institution).
UPDATE: this also got picked up by Daniel Drezner on his blog, and there's more discussion of the issue there.