Who let you in here, Per-Anders? And welcome! This gang needs a few good theorists with your practical experience.
> Cheap and good but they are 100 MHz transistors, according to datasheet at least. BC550C is 300 MHz.
May not be much different. As I'm sure you know (but others may not), Ft is roughly proportional to current over much of a transistor's range, so the different Ft numbers may just be different test conditions.
Anyway a transistor with β=100 and Ft of "only" 100MHz will have flat β up to 1MHz, which is higher than I can hear. If you are really needing a β of 500 (this amp doesn't need heroic β), then it is still flat to 200KHz. In a circuit with several voltage gain stages and overall feedback outside the amp, that can get you into stability trouble. In an emitter follower (even a 4-transistor follower), with strictly local feedback, it makes nearly no difference in the audio and above-audio bands.
For absolute best performance in the diamond buffer, especially if it swings into Class-B mode, Vbe/Ic linearity is perhaps the most important thing. This isn't often specified, mostly because the Vbe/Ic relation is fixed in the Silicon as long as you stay far below the transistor's maximum current rating. For hi-fi line amps delivering less than 1mA to the load, even 50mA parts are fine. At 600Ω Pro-levels, or headphones, think 100mA-500mA parts (and you probably need the bigger part for dissipation rating as much as for Vbe/Ic linearity).