Because the chance is ZERO, thus is worthy of promoting.
And if we keep asking for it, they may implement it in the expansion set.
Your logic is, at best, significantly flawed. Asking for a fundamental design change post-release is akin to asking the sun to not rise, or asking every single person in the United States to not drive for a single day. It won't happen, and it's unproductive to try.
You ask for customization. With a fluid skillset, you have far greater customization than if you lock yourself into a tree, because if you don't like a skill, you can swap it out for something else.
You say that a full respec is bad, because you want players to be penalized for making poor decisions. While there may be something to be said for investing players in their builds, what happens if/when builds become obsolete? The current system takes into account the fact that tweaks to the system will happen, and that skills will change. As a result, players are not penalized for design choices made by developers.
In this same paragraph, you make a point that having multiple difficulties is a bad thing, but then you talk about how much fun it is to re-level a class, just to correct for changes in skills, or skill choices made in a dearth of information. This is a re-hash of your "penalize the players" argument, but in any case, is a moot point in a system of fluid skillsets.
You end your post with a weak metaphor for people using the Internet to find ideal builds. In a skill-tree based system, min-maxing is encouraged, because generally, skills later are better than the skills introduced early on. If, for example, you know that you won't use any of the lv. 1-24 skills later, then why invest in them beyond the minimum required to get that lv. 30 frozen orb? To counter your argument, I'll use another food-based analogy: even the most delicious foods get tiresome after a time, so people tend to want a variety. In addition, the human body needs a variety of nutrients to survive, and most foods available do not provide everything we need. As such, variety is both desirable and essential - both for food, and for skillsets.
Now, since what your asking is not going to happen, and is futile to request in its current form, why not come up with something that adds to the current (not amenable to change) system? Here's my thought: Give each skill (not each rune within it) a counter. Once a player has 5 NV stacks, any kill that would generate a subsequent stack increases the counter on each equipped skill by 1. At specific cutoffs (say, 50, 100, 200, 400, etc), skills "rank up," and get better by some percentage (either increasing variable, numeric benefits, or possibly even adding effects) It wouldn't be so difficult to gain a rank or two in each skill, but if the difference between rank 10 and rank 11 great enough (coinciding with the number of kills it takes to get there), then players will be encouraged by the mechanic to keep specific skills. Again - skills would have counters, rather than specific runes. My thought on this is to allow for changes in skills due to developer changes in the mechanics, and to encourage some amount of experimentation. Now, not only are you investing in a skill, you're investing in the improvements that skill might get later.
You said yourself that a better model is to allow players to make decisions and tune their characters to improve their power, and this model would do so, without requiring a rewrite of major chunks of game code. As such, it is more like a compromise bill.