I scaled my A75s up to 58V output and 65 volt front end rails. The issues are component voltage ratings and dissipation. There are also some potential stability issues due to the way mosfet internal capacitances are affected by voltage applied. I haven't had an issue here (although poor layout the first time around

cost me a rebuild).
I used IRFD220/9220 for the input differentials. I ended up blowing a set due to oscillation ad replaced with IRF610/9610. I had intended it as a temporary replacement, but since they handle the dissipation better and I didn't notice a sonic difference they stayed. Over 65V you're really pushing the IRFD parts too hard, unless you reduce the bias current with potential sonic effects.
Another spot to watch is the current setting resistors for the zeners. You'll want to run them at around 5 mA, so change the values and dissipation ratings appropriately.
To get front end voltage higher than the output rails, the A75 uses a voltage doubler. Of course as the rails rise, doubling them means even more power dissipation in the regulator. A good way to minimize the problem is to use a 30VA or so 0-15, 0-15 transformer with one end tied to each of the main transformer to give you an extra 15V for the front end. Recitfy this and feed it to your regulators set for 10V over your output rails. Again watch the dissipation in the regs.
NP said that the A75 could get away with as few as 6 pairs of outputs. I take this to mean at standard rail voltages and using TO-3 devices. I use 6 pairs of TO-247s but my speakers present a benign load and I have PLENTY of heat sink. At 75V I'd use 10+ pairs and still ensure that you have plenty of heat sink - especially if you want to try to drive 1 ohm loads.
More than 12 pairs and you might run into drive current limitations since the A75 skips that stage. You'd need to increase the VAS current (and deal with that effect on stability) or add a driver stage.
Speaking of heat, I reduced the bias on mine to about 25 mA per device due to excessive room heating. I moved the Vgs multiplier to on top of an output device to help temperature compensation. It seems a bit overcompensated, but better than under. If you leave the Vgs where it is, stay with at least 100 mA per output pair, as NP suggests for thermal stability.
With 10 ouput pairs that will give you around 32 watts into 8 ohms class A, before you go AB - probably plenty for most applications. With 75V output rails that will be about 200W dissipation per channel including front end and regulator dissipation.
As for power output, if your front end is running at 85V, I'd assume that you can get to within 2 volts of your output rails. Assume they'll sag a bit or have a few volts of ripple. So you should be able to swing maybe 5V less than output rails. You can do the math from there, but your 1 KVA transformer starts to look a bit small to power two channels
