OK, turned out I had a similar design laying around in TubeCAD's Push-Pull Calculator.
I could not get a reasonable plate dissipation with the +/-180V supplies until I turned the grid bias WAY down. Assuming the tube will live 20% over RCA's plate dissipation ratings, I got -76V grid bias. P-P Calc gives these "answers":
The 100 Watts in 16 ohms is great. The 33 watts of heat at full power in a 13W-14W plate is not great, but may be allowable in speech/music use. The 4% THD is fine, can be much lower under feedback.
What really bothers me is the 744mA peak current in a tube rated for 125mA DC. I could accept 2X or 250mA peak in audio use, but 6X really seems unlikely to be good for the tube.
Linearity shows the reason for even 4% THD: the tubes are under-biased. Gain (red) is low at idle and increases on peaks. This is confirmed by the
waveforms which shows huge peak/idle ratio. The idle current can't be increased because we are allready at/past the plate rating.
A supply lower than +/-180V would reduce the too-high peak current and also the obscene drive voltage (135V peak!), and allow higher idle current for better peak/idle current ratio. This would of course reduce power, except that at 180V and 744mA peak, the tube life may be too to enjoy the power.
Changing to +/-100V supplies gives only 20W output, but a less unhealthy 362mA peak per tube and an easier 54V peak grid drive. THD is 1.4% 3rd, 0.06% 5th. Dissipation at full power sine-test is only 14.6 watts, utterly safe for speech/music work. +/-120V gives 450mA peak, 35 watts at 16 ohms at 2.3%THD, 70V drive, and only 18W dissipation in full-power testing.