I have been staying out of this and will immediately
withdraw from it, but I offer one observation, one claim,
and a couple references.

Observation:
I find it odd to watch historians of thought claim to settle
the issues of the usefulness of a macro-modeling framework
rather than serving as expositors and reporters of its use
and status within the profession.

Claim:
The AD-AS model is and should be taught in textbooks.  It
has much pedagogical usefulness.  It can be used to
illustrate stylized Keynesian, Post Keynesian, Classical,  
New Classical, and New Keynesian characterizations of the
macroeconomy, allowing students to (correctly) see macro as 
an area of theoretical dispute.  Additionally, it remains at
the foundation of all credible structural forecasting
models.  (At least I am unaware of any exceptions.)  The
IS-LM underpinning has only a bit less justification, at
least once the LM curve is replaced by an MP curve (as is
increasingly common these days).

Pointing out limitations of a 2 equation static model of
a complex dynamic economy is not particularly useful or
clever.  It has been done many times over the years. 
Discussing some of these  overviews seems like a reasonable
HET topic.  Pretending to correct the aberrant views of
people who make their living forecasting the macroeconomy
does not.  Instead it seems to recall the old question:
"if you're so clever ..."

To quote Leijonhufvud:
http://www-ceel.economia.unitn.it/staff/leijonhufvud/interview.pdf
        If certain institutional conditions -- such as
        fiscal balance over the medium run and
        a correspondingly stable monetary regime -- can be
        taken for granted, IS-LM will give you qualitatively
        the right answers to a set of important questions
        involving the system's short-term responses to
        various shocks or policy measures.  Then, IS-LM is
        a simple, handy way of thinking about -- and, of
        course, teaching -- how the economy responds. The
        trouble is, first, that the implicit institutional
        assumptions do not always hold and, second, even
        when they do, there are some types of disturbances
        for which routine use of it will not grind out the
        right answer. Learning IS-LM is easy.  Learning when
        it is safe and not safe to use it requires a lot
        more sophistication.

If historians of thought wish to discover what has been
pragmatically justified by people who actually do
macroeconomic forecasting, it is easy enough to find out.
Ask them, or read their stuff.  E.g.,
http://www.federalreserve.gov/pubs/feds/1996/199642/199642pap.pdf

Cheers,
Alan Isaac

PS Some points of reference:

Leijonhufvud, A. 1983. What Was the Matter with IS-LM?
In J.-P. Fitoussi (ed). Modern Macroeconomic Theory: An
Overview. Oxford: Basil Blackwell.

Gal=ED, Jordi, 1992, How Well Does the {IS-LM} Model Fit
Postwar U.S. Data?, QJE 107(2), 709--738.
(Answer: quite well.)