Here are my thoughts about bounded accuracy.
If hit points are the abstract portion of a monsters defense, which includes intangible elements like level and training that helps the monster survive longer in any given combat (as has always been the case since Basic D&D in the Gygax years), it makes sense to keep AC bounded for all of the other reasons mentioned in the L&L article that Rodney Thompson wrote. www.wizards.com/DnD/Article.aspx?x=dnd/4...
Having AC remain grounded in an absolute scale (armor/hide + dexterity bonus if applicable, plus any magical gains), maintains control within the system. If both hit points and AC scale (as someone else noted in another thread) then monsters grow in power at exponential rates which make it much more difficult for DMs to plan and run appropriate encounters. None of this means that a DM can't make monsters exponentially more deadly, but it does not assume it by default, much in the same way that the game does not assume that PCs will have X, Y, and Z magic items by Q level. By applying bounded accuracy, the core becomes much easier to manage at the default level. Individual gaming groups can tinker with the system to fit their tastes, and "taste" is the operative word. In any game, it is much easier to add then to take way. If the core is based on bounded accuracy, groups that don’t like its feel can add a number of abilities or even ad hoc bonuses to AC to make creatures more difficult to hit.
Giving monsters resistances and immunities becomes an easy way to make them more dangerous and unique without adding hit points. These abilities and many others can be used as a (modular) part of the design that will allow more people to play D&D the way they want to play D&D. If one group wants to treat a dragon as nearly invincible, instead of loading on more hit points, the DM can give it immunity to non-magical weapons, or resistance to weapons and energy damage (which effectively doubles the amount of damage necessary to bring the dragon down).
Allowing players to choose feats that augment bounded aspects of the game is also a possibility. There is already Shield Proficiency and Two Weapon Defense. Perhaps they can add Improved Shield Proficiency and Improved Two Weapon Defense, and other opportunities for DMs/Players to stretch bounded accuracy further if that’s what the DM/players want in their campaign. WoTC can even put an asterisk next to these types of feats that warn how using them may tip the balance of power in the game world (but an extra +1 or +2 earned through choice of trained feats probably wouldn’t break the game). They can do the same for Weapon Proficiencies. Perhaps there will be a feat that allows the player to specialize in one type of weapon or one particular weapon (a signature weapon) and gain an additional +1 to attack. Again, if it has an asterisk, new players and DMs will know that it may tip the balance of the game. These types of advanced training feats that may stretch bounded accuracy will make the core game play more like 3e, 3.5e and 4e. If people don’t like that type of experience, they can forget it and the game might play more like 1e or 2e.
Quite frankly, I think WoTC learned a lot from 4e. In 4e everything scaled, and that led to many complaints that PCs didn't feel as if they were getting stronger relative to the at level foes they were facing. Additionally, with scaled everything, combats took very long and there was still an arms race for PCs to gain more and more "to hit" and AC. With D&DNext, PCs will gain a sense of power and accomplishment by adding maneuvers and tricks or techniques that will give them more choices in combat, help build the narrative of the game, and allow players to minimize the desire to chase bonuses to offence and defense. I like that.