Note: this post is a bit convoluted. I'm sitting at work and don't have the time to thoroughly edit it; so it's more a stream of thoughts than a decent article. But as my schedule would allow me to work it out tomorrow evening, I post it anyway.
Your favoured approach - the scalpel - has its merits, but only if the thing you want to fix is pretty close to your ideal.
I was pretty much done with AD&D around 1999-2000, having shifted to other games and running the traditional group without much energy. A scalpel approach wouldn't have saved it for me. The 3e blowtorch was needed.
The concept of a continually supported version with an ongoing supply of material but with the core remaining intact and unchanged sounds nice. Yet even if they'd try to retain a (mechanical) constant version, WotC would have to update it stylistically.
What would today's teens think of the Easley covers of 2nd edition? Probably something like I think about stitched cushion sleeves: "Oh my...(shudder)"
The representation of rules in text has changed as well. The older versions had more or less isolated elements in spell and magic item descriptions. 3e started a trend of spearation of elements; feats, skills, prestige classes, they are all separate building blocks, each defined on its own. 4e continued the trend with the powers model. While those rules read terribly, the model has a distinct advantage: searchability. In the computer age you want to be able to find the definition of and rules for such an element easily and you want it to be presented comprehensively. A big contrast to old class or race descriptions, whihc didn't even have headings to organize the information.
From this follows, IMHO, that a constant corpus of a game can't be maintained indefinitely.
Another aspect you can't very well do with a scalpel is a change of focus. Let's look at the focus change from 3e to 4e. 3e used the model of an open construction tool and focusing on the possibility to add elements without limits to each part of the game. It didn't focus on balance and manageability of the resulting constructs. Balance could have been introduced by re-designing classes (scalpel approach), but manageability not.
Let's look at high-level buffing, e.g. Multiple buff sources with differing approaches to duration, keyworded stacking rules, and conditions made running such a combat a hell - for this humble DM at least.
4e introduced a focus change to the tactical combat encounter. (Likely) Actions became modeled in a rather rigid structure, durations were (more or less) unified, conditions were well defined on a very abstract level. This made the management of high-level combats much easier at the cost of introducing some "boardgame-think".
I don't think such a shift could have been introduced to 3e with the scalpel approach. The changes to the duration model alone would have forced WotC to re-write or even re-design an enormous amount of game elements.
NB1: Yes, I know, 4e combats aren't a breeze. Still, low-level fights with MM3 maths aren't as time consuming as many people seem to think.
NB2: The situation became worse with more and more interrupt actions.