I think there's a difference between inventing rules because existing rigid rules ain't cutting it and applying rules that require human interpretation.
As far as I can tell, there are three general methods by which rules may be implemented.
1. Individual, discrete rules for each occasion. E.g., because weather affects movement, discrete weather effects should have explicitly-specified impacts. Because ambient temperature affects health, definite temperature ranges should be defined and clearly specified.
2. Universal, abstract rules which apply to genuinely all situations. E.g., Fate uses aspects for...everything, as I understand it. Or certain forms of the "saving throw" idea, since "roll to avoid <Bad Thing>" is pretty abstract. A mechanic meant to be sufficiently abstract that there is almost nothing it cannot be applied to, though it may require human effort to translate that abstraction into a concrete application.
3. What I call "extensible framework rules." They aren't the individual specificity of type 1, nor the (near-)absolute universality of type 2. Instead, they are defined for particular
sets or
ranges of situations. Those sets can be infinitely large, without being "literally (almost) all possible things." An extensible framework is somewhat abstract (since it can range across anything within its relevant set), but easy to translate into concrete things (because it does not have to apply to dang-near everything.) Unfortunately, this means such things can also be the worst of both worlds, requiring too much specificity to be generally useful, yet too abstract to make clean, obvious sense of.
Most systems use a mix of types 1 and 2. Fate leans very very hard on type 2, while 3rd edition D&D leans very very hard on type 1.
Now,
separate from that consideration, there is the question of whether the rules are comprehensive, or not. A system going for type 1 is going to struggle mightily with
absolute comprehensive coverage, because you would need infinitely many rules for that. However, as long as the places not covered are rare and generally not very important, it may be possible to get
effectively comprehensive coverage with only finitely many rules....but it will usually still be a
lot of them, which is a weakness.
A system going for type 2 has it easy in this sense, because a truly universal mechanic is
designed to be comprehensive. However, as noted, the closer you get to true universality, the less the mechanic actually
tells you about how to play and what to do. You are offloading specificity to the end user. This may be a huge benefit, but sometimes questions of that nature can be real tricky and it would be nice to just have a clear, straightforward answer. If an Aspect is genuinely
anything that could be relevant, knowing that something is an aspect doesn't really tell you anything except
that it is relevant.
If someone is approaching things from a type-1-centric view, then type 2 looks like constantly inventing new rules for every situation. Conversely, coming from a type-2-centric view, type 1 looks like a byzantine morass that
still needs its edge cases papered over anyway. Type 3, being midway between, may be either a refreshing approach to thorny problems or a frustrating lack of commitment to worthwhile goals.