With RPGs I think the most important thing is to have clear design goals and know your audience. I also think not limiting yourself to just one approach or one design philosophy is helpful. I tend to prefer streamlined systems with unified mechanics. But I think there are also going to be times when being inefficient and less unified can be very effective. There are advantages to both approaches. One of the obvious downsides of a unified mechanic (what I believe you called consistency triqui) is your core mechanic may not be the optimal choice for all aspects of the game. It might be clunky to have different mechanics for different rules (like in AD&D with different die rolls for non-weapon proficiencies, attacks, initiative, thieve skills, etc), but the one advantage is you have way more control each of those aspects of the game.
While I think overdoing any game design theory is a flaw (ie: shoehorning some rule to fit into other rule you already have for consistency), I tend to think that, given two approachs that both work, the one with consistency is better. Using as an example, 3e: While it *might* be a mistake using the d20vsDC for, say, damage, it *has* merit to build all skills, attacks, and saving throws, under the d20 system.
My point with the thread is that, before I read the Rule of Consistency for the first time, I wasn't even aware of this. I wasn't aware, for example, that D&D stats and skills weren't consistent (ie: 3-18 vs ranks), while World of Darkness stats and skills are (1-5 dots). This does not mean whenever I build a game system I'd make stats and skills consistent. Sometimes I'll do, sometimes not. But the fact now I know this exist, I'll
look at it.
And yes, I'd also like to make use of different approachs, depending on what I try to acomplish. That's the point: to use different tools, those tools have to be known. In modern science and techniques (be it economics, engineering, software development, whatever), having precise definitions of techniques help
a lot to organize work, avoid flaws, and increase efficiency.
Knowing the difference between Realism, Believability and Verosimilitude for example, helps to clean out rules for a given desired effect, just like understanding the difference between Accuracy
and Precision
would help to build the right firearm for a given military job or the difference between effectivenes and efficiency in economics. They might sound pretty similar, but if you build careful definitions, it's obvious they aren't.