Michel, I think both kinds may have existed, but as you stack more balls, the four sided is more efficient of
quantity. Either system will work, and regardless, the shrinkage issue is real. Brass has about twice the coefficient
of expansion of cast iron. Most bimetal strips for thermostats have been made with iron and bass strips fused.
BUT, regardless of geometry, it could happen. And maybe only the four sided ones were used. I do know of
at least ONE triple stacking setup (monkey may be buried here) The is the Tsar's Cannon in the Kremlin in Moscow.
The story is that the Tsar (like any good Russian) JUST HAD TO HAVE the biggest cannon, so he had this one with a
one meter bore made up. You will note, that unlike most cannon of the era, the barrel wall thickness is essentially
constant, not thicker at the breech end. A rumor has it that the four balls are 1 cm larger than the bore diameter,
just in case some latter day fool decided to try to actually shoot it. I wish I had an accurate caliper at the time.
Here is a square stack.
The one question I still have, is whether one can find a reference in ancient literature where the cannonball storage
plate/rack is called a "monkey". This one has always 'seemed like it is true' but I will not be totally certain
until I read a book somewhere written in 1840 or so about storing their cannonballs "on a monkey".
I personally call this one 'credible, but unproven' at this point.
Bill