It think I see what you mean. There are two kinds of linearity in play. For different frequencies you have linear addition of the powers at each frequency. At the same frequency, you have the power of the linear combination of the two waves. It all boils down to the orthogonality relations for trigonometric functions.

Consider a voltage wave, \$V \ = \ A\cos\omega_1 t \, + \, B\cos\omega_2 t\$, if you are to compute power, you must integrate its square. Squaring: \$ V^2 \ = \ A^2\cos^2\omega_1 t + 2AB\cos\omega_1t \cos\omega_2t + B^2\cos^2\omega_2t\$

If \$\omega_1 \ne\omega_2\$, integrating over the period makes the middle term go away, and you are left with the powr of the individual components. There is linearity.

If \$\omega_1 = \omega_2\$, you have \$V^2 \ = \ (A^2 +2AB + B^2)\cos^2\omega_1t \ = \ (A+B)^2\cos^2\omega_1t\$ . The individual waves do not contribute their power, but interfere and you get the power of the *linearly combined* wave.