RAM heatspreaders are-in the main- a holdover from DDR2 technology, where 1.8v was default JEDEC spec and 2.0 -2.35+v was the norm for overclocker RAM (either DDR2-800 @ CL3 or 4, or DDR2-1000+ @ CL4) and heat dissipation from the IC's was an important consideration especially once you exceeded ~2.2v
Modern DDR3 modules tend to sit at an upper limit of 1.7v -with many binned IC's happy to OC at 1.3-1.65v- with the requisite lower temp output. Also bear in mind that modern RAM is very tightly binned -that is to say that even with aggressive voltage and relaxed latency, modules
generally wont reach the next speed bin (i.e. 1600 wont be pushed to 1866, 1866 wont go to 2000, 2000 wont go 2133 etc.) regardless of voltage -and of course with voltage limits now becoming widespread as a function of the CPU there is less voltage overhead to be used, whereas many DDR2 modules would accept out-of-spec voltage in exchange for tighter timings and/or increased bandwidth (Example being the Crucial Ballistix in my secondary rig are nominally DDR2-800 @ 4-4-4-12 - throwing 2.3v at them on a good board (strong northbridge) will get them to DDR2-900 @ 4-4-4-12 or -1150 @ CL5).
So as far as your poll is concerned, either option is valid to a degree. DDR2 @2.0+v is a definite yes unless you have server grade airflow. For DDR3, any heatspreader is going to help draw heat off the IC's- but in the main it's only going to amount to a few (low single digit) degrees-
unless you're caning it, so any cooling is good. From a personal viewpoint, I tend to steer clear of tall/tall finned/watercooled heatspreaders -they are more affectation than performance. They also tend to limit air cooler choice and make an effective air dam that disrupts airflow over the motherboard power regulation/mosfet heatsinks- neither of which is particularly desirable.