Personally, I find the cathodyne the most technically appealing in a general sense. It is very simple and transparent, and balance is determined by the resistors rather than the tube being used. For any non-class-A load, however, they need a proper differential amp as buffer (Williamson style). Dynaco circuit doesn't cut it IMHO for class AB.
The long-tailed pair with a CCS cathode load, however, is hard to fault and my next amp will use this approach.
I also like the plate-follower paraphase used in the Lafayette 550 amp - in theory it can have lower distortion than a cathodyne and has twice the gain - but introduces more variables in terms of balance and phase shift.
I'd consider an input transformer, but the cost is too high for truly good ones (high-nickel Cinemag, etc.).
The only one I am not comfortable with is the conventional paraphase - since it has the most variability in balance. (Sorry Andrew! - certainly they can work, but need a balance pot to really get it right - I trust you'll pull it off somehow
)
