It's not a new fad. Ralph Glasgal has been promoting DSP crosstalk reduction for a long time. Polk, Carver, Lexicon and others have tried to tackle the crosstalk problem over the years using analog methods. It's a real problem, a real shortcoming of 2 speaker stereo reproduction. It's not intentional or desirable in any way, although some may develop a taste for it. 2 speakers is the simplest stereo possible and it beats mono, so that's why it became a standard. It has nothing to do with any inherent audio superiority, although with crosstalk reduction it can get so good on properly made recordings that it's harder to argue the need for multi-channel formats. Still, I think well done up-mixing of stereo recordings to 5 or 7 speakers located across the front of the listening area has some great upside potential, especially for creating a huge sweet spot. Purpose made recordings for this kind of setup would even be better, but that's not likely to happen.
Adding crosstalk to headphones may make some sense, but ideally you would not want to do it with the same time delay that happens across your head with stereo speakers. If you do so, you'll introduce massive comb filtering to the signal for center panned sounds, just like speakers do to your ears. Now that I say that, I can think of a way to add time delay without causing comb filtering to center panned sounds - I think I'll have to try that!
I've not tried the BACCH DSP yet, but I'll vouch for crosstalk reduction. From the reviews I've read of BACCH it sounds similar to my own experiences with using a physical divider or my own, not so powerful channel mixing methods. The timbre of center panned sounds is very nicely improved, as is the overall sense of acoustic space. It matters more on some recordings than others. I never heard anything that sounded worse because of it, although I'm just reducing crosstalk, not creating the extreme degree of left right separation that BACCH DSP can.
