In '99 we experienced a head failure (BGMR with turbo) that torched longintudinally through the combustion chambers, blew a hole through the timing cover, burnt the cam belt idler pulley bearing, and finally blew a water hose off at a firewall nipple. Our rectangular tank held, though it did bow the sides and top. That tank was made from 16 guage mild steel, with a single central baffle. We used a pressure cap, with relief into a catch tank. When the pressure spiked, the cap couldn't relieve fast enough but that's what happens with catastrophic failures.
The pressure cap question is a hard one to answer; if you raise the pressure it's harder for the water to boil. I now run both radiator (w/ pressure cap and recovery tank) and large tank on my pushrod bike, but I run reverse flow (into the heads, and out through the block...I can make water heat but not oil/block heat). I run the radiator immediately after the block (before the cooling tank) because a radiator works best when temperature differential is as high as possible. This has worked perfectly (learned it from the guys at TRD).
A perfect example is Fuel Cell cars, which need a LOT of real estate for huge radiators. Not because they make that much heat, but because the temp differential is too small (compared to ambient air temp), yet the stack needs temperature controlled accurately.
I forgot that lesson in 2010, running a pair of radiators in series (without a water tank). I was very little better off than a single radiator, because the second radiator did not contribute much calorie absorption while adding additional restriction to the water flow! Ooooops.
food for thought, I guess.
JimL