I must say, considering the number of years keyswitched articulations have been around, it is hard to believe Steinberg beat Cake to the punch on this. I do not use Cubase so I can't swear to it's functionality, but here is why Sonar needs a feature similar to this.
Consider a violin passage that needs to start in legato, go to tremolo, then perhaps detache. If the vehicle for that is a multi-instrument where each articulation is called up by a program change you are golden. Sonar has program change searchback and will "remember" what articulation is currently expected (it looks back into the even list to find the most previous command and invokes it just as you hit play, from wherever you are at in the score).
But if the vehicle for that is keyswitching, then, if you stop when the instrument was playing tremolo, then move your cursor to somewhere in the middle of the legato passage and hit play, you will be gritting your teeth as you hear your lovely legato coming out in tremolo... because Cake doesn't searchback to keyswitches because it doesn't know that they aren't legitimate note events.
That is the crux of the issue... developers used a crutch for program changes (for legitimate real-time performance needs) and circumvented functionality that had been designed to handle program changes in design-time... without a corresponding accommodation from sequencer manufacturers to allow keyswitching to somehow have searchback behavior. What it appears Cubase has done is to create a mapping applet that lets the user define articulation maps that associate specific articulations with specific keyswitch notes, much the way you can currently define instrument maps, or drum maps, now in Sonar.
I could really really use this feature (assuming it would searchback like program changes do). It's a royal pain in the ass to be constantly looking at my cheat sheets to find the appropriate keyswitch for the current instrument I'm working with (keyswitches are all over the place and non-standard, as they have to be, because of the different playable note ranges of various instruments.
I was recently thinking about solving the problem with KSP scripting (Kontakt 3) but program changes aren't handled events, so no go there. If they were you could add a script to an instrument that intercepted a program change and called a note event based on a lookup table.
For the first time in my muscial life the idea of looking at another sequencer has occurred to me... frankly it's appalling to my mind, but there it is.