There's plenty of posts regarding the the forumlas. One that is typically referenced is valk's data that he collected in Beta 3 of FFXIV 2.0. I know others probably have done their own share of theorycrafting as well, and data seems to be spread throughout the community. For the most part though, valk's forumla is fairly accurate, but is usually off by 1% in terms of healing (so if you're Cure heals for 1000, valk's formula might predict 990). 10 hp isn't that much of a difference, but that's about how much increasing stats does to your single Cure.

As for collecting data, if you know much about how the data is generated in the game, you'll realize you only need to collect very little to determine the trends. All heals have a variance of +/- 3 percent, so it's easy to establish a median by looking at the range of Cures. Depending on the cure strength, it's possible to know when you've obtained the exact value of healing. So once you realize you probably only need ~100 Cure's per data set, you really don't have much work. Secondly, you mention that parsers aren't accurate. The fact, in general, is true, but you have to also realize, why they aren't accurate. All parsers estimate DoT and HoT because of the way they are accessible to the parser, but every single heal except Regen and Medica II can be perfectly collected by parsers.

Because your assumptions can be easily changed. It's not nearly as problematic as you say regarding your weeks of coil data. Healing isn't as interesting as DPS, since the point of healing is to match the incoming damage with outgoing heals while minimizing mana usage over the same period of time. DPS try to maximize damage in the period where the boss is alive. The terms for determining BiS is very much different.

And to answer your last question, I have an in-game macro that casts Cure x8 and it's pretty easy to hit while watching TV if I did need to collect data.