Lots of interesting discussion here.
Here's the way I see it. It's standard to make the distinction between "player knowledge" and "character knowledge." What this discussion also shows is that it's also necessary to make the distinction between "player decisions" and "character decisions."
Ideally, in order to preserve consistency and prevent accusations of "metagaming," we want to preserve the following conditions:
1. "Character decisions" are not made using "player knowledge." It's also important to note here that if a particular decision is a "player decision," then the fact that that decision is there to be made is "player knowledge." For example, if using an action point is a "player decision," then a character could not make a decision that relied on being able to spend an action point at a particular time in order to work, since he wouldn't know that he could choose when he spends action points.
2. Characters can make logical deductions. In other words if "A" is character knowledge, "B" is character knowledge, and "C" can be inferred from "A" and "B", then "C" should be character knowledge.
One way of solving this is to make all game mechanics information and decisions "character knowledge." This is the position I advocated in my original post. It is perfectly self-consistent, although some players may not like the idea of characters knowing this kind of information, and would prefer another solution.
Another solution is to make certain things "player knowledge," such as power recharge timers. The only problem is that in order to preserve the above conditions, you have a "cascade effect." If you want power recharge timers to be "player knowledge," but characters would be able to infer the existence of power recharge timers from the results of their decisions about power use, then you have to make all power use decisions "player decisions." And if power use decisions are "player decisions," then pretty much all in-combat decisions are "player decisions," because most in-combat decisions are geared toward trying to use powers more effectively. And so on.
Of course there's always the option of weakening the conditions, like just agreeing that characters won't ask too many inconvenient questions (weakening condition 2) or being okay with some "character decisions" made based on player knowlege (weakening condition 1).
-----
Also, now that I think of it, this discussion has illuminated an issue of the debate that I've always had trouble understanding: why people complain about 4e powers' recharge timers, forced movement, specific rule-based effects (like 'granting an extra save'), and effects that are not based in "reality", while they didn't complain about the same things with 3.5e magic.
Here's the deal. Suppose that we want characters to be in control of their use of powers in combat. Some powers and abilities require characters to know about game rules, while some don't. For example, a basic attack doesn't require the character to know about game rules. He just attacks, and if he hits, does damage. The character doesn't need to know about, say, attack and damage modifiers in order to effectively use a basic attack; they're just there.
However, consider a power that can give an ally something when "an ally receives a condition that a save can end." (I believe there is a power that has this trigger.) Unless this power just goes off without any conscious control by the character, to use it implies that a character must know what a "condition" is, and tell the difference between an "until end of next turn" condition and a "save ends" condition, so he knows when he's eligible to use it. Similarly, a power that allows you to "recharge an encounter power of your choice" would imply that a character knows what an encounter power is and which ones he has.
So what does this have to do with the issue mentioned above? In 3.5e, wizards did have to know about things like spell levels, prepared spells, level-dependent durations, which immunities worked against which spells, etc. in order to use their spells most effectively. But this made sense, because wizards are traditionally depicted as gaining their powers through study, so it made sense that they would study magic and know about how magic works, which would include these concepts. Thus wizards could conceivably talk about spell levels etc. in character, and it would be the D+D world's equivalent of computer jargon: understandable only to those who study it.
On the other hand, 4e gives these powers to all classes, even ones that aren't depicted as particularly intelligent or gaining powers through study. An analogy would be an athlete trying to catch a ball - he doesn't have to know about the physics behind how his joints move in order to do it. But if an engineer were to build a robot to do the same thing, he would have to understand the physics in order to program the robot. But 4e makes all classes like the engineer - they have to understand the rules in order to use their powers effectively.
Or, to sum up: In 3.5e, wizards and other magic casters are the "geeks" of the game world, and people accepted that because that's part of the idea behind a wizard. But in 4e, all classes are "geeky" in the same way the wizard was.