Actually, the level of ocing i am doing and have been doing is and was quite achievable on air. I was using on my old rig the Thermaltake DuoOrb cooler, and had planned on migrating that to this new rig. (I won't get into the hassle i had finding a bracket that would work with both the lga 775 cooler and my new lga 1155 board, but i did) But the ram i bought, the heatsinks are too high (Corsair Vengeance with 75mm heatsinks) and the cooler wouldn't fit.
After trying a few other coolers, i then had to consider water. so it was a necessary step just to have reasonable cooling and still use all 4 ram slots (4X4GB). I had considered water before, just because i can, but found the DuoOrb to be a fantastic cooler and on my old, but still trusty Q6600 i managed to post on 4.0GHz (2.4GHz factory) in my Antec 900 case.
Im in the same case now with new build and have had the i5 to 4.4 just to see if i could, but dropped it back to factory, and have been running that way the past month.
Anyway, i haven't really answered your question. I mess around with ocing for fun, but have zero need to do it, nor do i need the water cooling other than it was the only suitable cooling solution, form wise. I haven't built so many rigs myself, perhaps 10 to 15, but this is the first time i have ever messed with water. It's a great talking point, none of my friends are too computer literate (hardware wise at least) and had never seen anything like it, but if i could have used my DuoOrb, i would have!
I used to love Toms, but since the site redesign i haven't been as much a fan. I don't really care for the new look or the new presentation of the charts and such, but it's still an invaluable resource when looking to update ones rig.
Thinking along your line of suggestion Jafo, i will take apart the rig this weekend and reroute the piping and relocate the res/pump/rad (it's an all-in-one unit). It's not the most efficient route i have going, since i routed with aesthetics in mind, and i see where i might have a slight restriction, though there is certainly no evidence of that.
Since the earlier post i have been running prime95 and been checking the temps when in sync with noise occurrences and i can't say i see any pattern, temps are stable, both under load and at idle.
I'll keep scouring the interwebs and i hope more help will surface here too!