On AMD platforms RAM does have some impact, on Intel however people should simply stop taking all those RAM tests seriously.
The difference in performance that benchmarks show is extremely artificial and it’s largely irrelevant outside of the overclocking for the sake of numbers game.
The reliability argument is another joke played on customers, in day to day activities on RAM that isn’t aggressively overclocked, and you don’t get your overclock on Ivy or Haswell from the base clock, you’d be very hard pressed to tell the difference between the cheapest plain sticks from a reputable vendor and the absolute top of the line ones costing time and half as much.
Stick cooling past a certain point is also a bit of a joke. Those massive flimsy constructs on modules like the vengeance make absolutely no difference in example.
No heatsinks at all with good flow and making sure there’s no stale pockets will be a million times better than some fancy, thermodynamically laughable highly turbulent thin fins that sit in a pocket and almost touch your CPU heatsink.
Get 1866 to have that little bit of elbow room for the last stretch of performance (IE: a bump from 100 to 105), and so that if you really, really can’t live without it you can drop one of the CAS a notch, and go for low profile, so that you don’t have problems with coolers or installation.
Personally I’d go for 2x8 if all you want is 16GB, so you have room to spare if you decide to upgrade, and better resell value.
Ares 1866 C10 are perfectly fine. Well built, low profile, rock solid, and priced right, not to mention massively adopted on Ivy.
As for the cooling, I still have to see ONE picture or article about an H50 that leaked other than the couple that have been doing the rounds that were patently over-torqued cabling or mis-installation, something you would really have to make an active effort for it to happen.
But that’s up to you. Do bear in mind it’s an absolutely irrational fear statistically speaking, fans are a lot more likely to fail with age (and trivial to replace) than a submerged pump or a cable seal are. Cheaper and worse built closed loops operate in more aggressive conditions than that for decades before failure.