As a motor is not sensitive to voltage you can use PWM on it without exceeding any peak rating. The heat will be spread over the thermal time constant which is long compared to the PWM frequency.

So yes you can.

I beg to differ. In the absence of substantial current smoothing due to motor inductance, the power dissipation will be greater with PWM compared to a steady voltage.

Consider a 10V motor with a resistance of 10 ohms. When stalled and fed from a steady 10V, the power dissipation will be 10V*10V/10ohms = 10W. Suppose we now run it from 20V with 50% PWM. For 50% of the time the power dissipated will be 20V*20V/10ohms = 40W. The rest of the time it will be zero. So the average power dissipation has doubled to 20W even though the average torque is unchanged.

Now suppose the motor is running but loaded, so it is producing a back emf of 5V. When driven from a steady 10V, the power dissipated is (10-5)V*(10-5)V/10ohms = 2.5W. When driven from 20V with 50% PWM, for 50% of the time the power dissipated is (20-5)V*(20-5)V/5 = 22.5W. The rest of the time it is zero (the back emf is in the wrong direction for the flyback diode to conduct). So the power dissipation has increased by a factor of 4.5. In fact the motor torque will also be higher, because the average current has increased from 0.5A to 0.75A.

So I maintain that driving a motor with above its rated voltage and using PWM to limit the average voltage is only safe if the combination of motor inductance and PWM frequency is sufficiently high to substantially smooth out the current.