How Much Should We Worry About Ability Scores?

I know I'm already working on two other multi-post topics here, but I thought I'd take a break and draw attention to What's the Point of Ability Scores? (Part I) over on the Grognardia blog. James Maliszewski is diving into the topic of what ability scores used to mean vs. what they mean at later points in the evolution of D&D and related Class and Level Exploration Fantasy (CLEF) games. The D&D community went through several stages:

1. referee rolls ability scores, player chooses class afterwards
2. player rolls ability scores, then chooses class afterwards
3. player rolls ability scores, chooses class, and adjusts scores bases on class (point swap in Holmes, etc.)
4. player chooses class, creates ability scores targeted to fit that class (AD&D 1e Method V, point buy in later editions)

These stages of development sort of match up with changing views on the importance of ability scores:

1. guidelines only, few iron-clad mechanics tied to scores (3 LBBs OD&D)
2. some modifiers for high/low scores, gatekeeper function for certain classes (post-Greyhawk OD&D, early AD&D 1e)
3. modifier/gatekeeper functions + roll under rolls for non-weapon proficiencies (late AD&D and Classic D&D)
4. modifiers to "universal" game mechanic (D&D 3e and later)

In other words, the more important ability scores become, either for getting the class you want or just plain survival, the more players are going to want more control over their ability scores, either through guarantees that they can get at least one high score or through actual point-buy.

Since I've made no secret that I prefer using ability scores as guidelines and have been removing modifiers from my own gaming as much as possible, the low numbers on both of the above scales are my sweet spot. Although as I mentioned in a comment on Grognardia, I'm pretty committed these days to the idea of "either roll 3d6 in order, or just pick your scores and let's move on". Some of my ideas on that I've covered before here.

• Let 'em Pick
• Let 'em Pick random table (PDF)
• Just Pick a Class Already

I'm not sure where that would fall in the development of ability score generation methods, although I like to think of it as Stage 5. Once I realized that all ability score generation methods were basically about getting as close to the scores you want without the GM and other players thinking you are cheating, then really everything other than "I'll take whatever scores I get" is some form of "I'll try to get the scores I want". It's just easier to cut out the dice rolling or point distribution tricks at that point, pick the scores you want, and move on. And if the GM is not interested in being a dick about what players can do, there's not much reason to worry about cheating on rolls, is there?