Well, I think I'm the only one on a few different forums who has any experience with this monitor. And being in such an exclusive group means no one can really help you when you have issues. Wells Gardner's support is pretty much nonexistent so you have to come to grips with that as well.
To put it simply, my D9500 is broken. For whatever reason, it will not work with a composite sync source as it is supposed to. It's going to require me to send the chassis in for repair/replacement. That's not fun. The whole ordeal hasn't been very fun and took quite a while to get to a point where I can hopefully get them to repair the stupid thing. Billabs is selling a model with similar features, the BL19C91T. I can't vouch for Billabs, but the general consensus seams to be "Stay away from Wells." Wish I had listened.
Anyway, to address your questions. The 25khz games should display perfectly, although I haven't actually tried any on my monitor. Games do look fantastic in windows, as do games at 15khz. As far as having to adjust horizontal/vertical size based on the game, well, I'm pretty sure you're SOL for the most part.
I've had the monitor since march of 2006 but since I haven't completed my arcade cabinet, I haven't used it much. However, when testing the monitor while connected to my pc I have tried a few games. At least with advancemame, changing between games that use different systems causes the same type of phenomenon that you wish to avoid. Basically, if I adjust the monitor to play a game based on the neo geo hardware, when I switch to a game based on cps2 hardware, I get black bars at the top and the bottom of the screen as the two systems run at different resolutions. It's annoying, but I don't think it's really the monitors fault. That's just the price you pay for running the game in a pixel perfect fashion.
I suppose you could force the games to display at a certain resolution and then you'd never have that problem, but the games wouldn't technically be running in native res anymore. Then again, I could be wrong.