I'm confused, you said halving the duty cycle doesn't halve the power, then said that when the PWM halves (from 100% to 50%) the power halves (4xPrated to 2xPrated)… I'm definitely not trying to be difficult I just think the math says otherwise.
Let's establish some rated conditions. At Vr=12v, the heater draws Ir=Vr/R where R is the (more or less) fixed resistance of the heater. The power Pr=Ir*Vr=(Vr^2)/R.
If I have a switch driven by Vin and pulse the switch with a 50% Duty Cycle, the average output voltage is 12v. Because we only care about the long term effects (relative to the PWM frequency) we can use the average value in our calculations.
So at 50% DC (duty cycle) with Vin=24v, Vpwm is DCVin=12v and because the heater resistance is broadly fixed, the heater Power is (VpwmVpwm)/R, which is also (DC^2*Vin^2)/R.
So doubling Vin means DC^2 needs to reduce by a factor of 4, which means the duty cycle is only reduced by a factor of 2.
Let me know if I got something wrong, but that's why I think you'd limit the PWM to 50%. The non-intuitive part comes from squaring both voltage and duty cycle in the power equation.