how is CHS's performance once the database gets large such as 400MB+?
That's a good question, but i don't know the answer.. If there is interest we could find out -- I could write a function that generates some random clip data and we can see how it performs.
My suspicion is that, assuming the clips are in different "folders", so that it's not trying to sort and search 400mb of data at a time, that CHS will perform pretty well, but that's just my guess. It's also quite possible that it performs very similarly to ClipCache... A factor involved here is how many clips there are (a relatively small number of large clips will perform better i would guess than a million clips).
What I will say is that CHS does try to "help" you not accumulate 400mb worth of clip data, by employing configurable options that will automatically move default (non-favorited) captured clips by default from new to old to trash groups.
As for importing, CHS doesn't have a way to import clips (though it can export them). I would be willing to write a simple import function if you can get your data out of your existing program in a fairly clean and simple format. The easiest would be one line per clip, with a line being just the clip text, no extra data like date, etc.