Perhaps the reason there is such diversity of opinion on this issue is
because of the complexity of making an optimum determination; there are
a large number of variables.  To mention a few; flow rates, tube
thickness, tube wall thickness, fin density, thickness and design, air
speed, inlet diffuser design, blah, blah . .  .  So there is a tendency
to make qualitative and intuitive judgments which ignore all but a few.

The criteria are size, weight and drag.  The Nascar and other racing
folks have done a lot of optimizing that applies for us.  I think there
is good reason why most the race radiator vendors make rad cores in the
2 - 3.5" thick range.  I still have the plots from some fairly extensive
studies that Fred Moreno posted on the "other" list a couple of years
ago.  Using values for the rad design that are "typical"; his studies
show a fairly broad optimum for core thickness between about 2.5 and
3.5".  

Perhaps one factor that Tracy's argument may have slighted in the
thicker cores is the decreasing efficiency of heat transfer as the temp
difference between surface and air temp gets lower going through the
rad.  Thicker may be fine if the fin density is a bit lower the fins are
a bit thicker.  It's difficult to say that there is one right way
because each installation is different; so it may be best to go with the
body of evidence that says somewhere in the range of   2 - 3.5 inches.

One thing is for sure; you have to have the pressure recovery needed to
overcome the pressure drop in the core.  So if you are going to hang the
radiator out in the free stream air, it better be very thin.  If you
have an effective diffuser design the slows the air by a factor of 4 - 5
then be happy with your 3" thick core.  If you want to minimize drag on
the plane, have an outlet duct that efficiently accelerates the air back
to an area of about 1.5 - 1.8 times the inlet area.

FWIW;  Al Gietzen