That's an interesting looking device. Have you actually tried using one? Or anyone else have any experience with one? I'm curious how well something like that would actually work in practice.
I've tried using a Kinect for motion capture purposes, and found it to be extremely noisy, and the limited viewing space really restrictive. It also would glitch out massively if you ever self-occluded yourself and put say one leg in front of the other. I have hopes for Kinect 2, though it'll suffer from the same self-occlusion problems. Though you can alleviate those somewhat by using 2 Kinects at different angles (haven't tried that myself, so, I don't know how well that really works).
Plus of course, some actually has to be able to physically perform the motions you want to capture...
You can add a delay to every action (or at least to slower actions), giving you a natural opportunity to evade it (provided you have enough time).
Say you are going to act next in 5 ticks of game time, and your opponent begins an attack that takes 6 ticks to land. Your turn will come up before the attack connects, letting you either move out of the way of the attack or perform a guarding action. Hitting someone is then a matter of either timing your attack right so they don't have time to evade it, or positioning yourself such that they cannot physically move to evade it.
Then like a fighting game, you can have different amounts of hit/block stun on moves, or punish someone for missing an attack provided you can move and attack while they are recovering from missing their attack.
This would work best in a team based game, otherwise it would be too easy to just pick the optimal move at all times and avoid everything. You'd also want area of effect and other such wide angle attacks to limit evasion and be able to be used to position your opponent where you want them. Same with attacks that push your opponent around, being able to shove a blocking opponent into the path of another of your character's attacks would make for an interesting mechanic.
It also lets you put weight on spending time to progress your character. Crafting/making an item will take time (in which you could be doing something else, or pay someone to make it for y ou). Learning a new spell/skill would take time.
You'd want to give some leave way in the exact time that events occur in. Otherwise, it will be almost impossible to see any of them. You'd also want to give the player some knowledge/idea of when events will occur. Reports of enemy army movements would let them know about when they would have to arrive to help defend a location. Also if the player is in a location where an event is going to happen and decides to rest/sleep/wait, interrupt that when the event happens instead of skipping over it (and give an option to wait for it, when the player knows in advance when something is going to happen).
The main problem though, is the sheer amount of additional content you have to make. In a more linear game, if the city isn't sieged until you arrive to save it, then, you never need to create the destroyed/conquered version of it.
This will also easily lead to an "unwinnable" game, as odds are, the world will be destroyed without you. Course, for a shorter game, this wouldn't be that bad, especially if you want it to be replayable (you'll have a better chance of that, if restarting the game to try a different sequence of events isn't that huge of a time investment). You'd probably also be better served by a less "epic" scope than saving the world from destruction, where the consequences for not arriving aren't quite as dire.