If you stay with indirectly-heated dual triodes, here are the limitations:
The high voltage is regulated to 300 volts, so that is the maximum plate voltage rating for the regulator. Most dual triodes will meet that. (The driver itself operates at around 175v.)
The standard setup has about 3.4 mA through the plate of the driver. The 5670 saturates (zero bias voltage at 3.4mA) at 55v from the published curves, so the peak output is +/-125v so in the stock Paramount with 70v bias, the driver operates at 56% of its peak capability. This margin reduces the driver's distortion contribution to the total, which is one of my design goals. To maintain this situation, you need a tube with similar plate resistance, unless you adjust the plate current to re-optimize.
The bias of the 5670 is about 4 volts, and as designed the bias (from the 431) cannot be adjusted below 2.5 volts. In practice this means you can't get away with a mu greater than around 40-50 because the bias would be less than 2.5 volts. A lower mu is pretty easily accommodated, and the obvious candidates are the 6SN7 and 12AU7. These tubes will work in the stock design if the bias adjust trimpot is changed to 20K from 10K ohms.
Since you are running the amp at lower 300B current, there is more current available for the driver, so tubes with moderately lower plate resistance are possible if the board currents are adjusted. But the board's small heatsinks can't take a lot more current. I would be reluctant to go much lower in current because of the Miller capacitance of the 300B. A current as low as 2.0mA should work (this would be about right for a 6072), though I am not confident it would sound as good.
I don't have detailed designs for other tubes at present.
