Just as a qualitative argument, if you must avoid using channels that are already in use, and the SNR is poor, you need to gather more samples. Fading will create times when you receive no detectable signal at all, and yet the channel is in use.
This will matter more in cases such as TV spectrum white spaces, than in cases such as WiFi spectrum sharing.
In the TV case, the priority signal is likely to be extremely weak. The TV transmitter will often be much further away from a legitimate user than any opportunistic transmitter trying to use white spaces. And even though the TV transmitter is more powerful than anything an opportunistic user might have, propagation loss, including terrain effects, will easily drop the TV signal strength to very low values.
In WiFi, the signal that matters most to the users of the WiFi is likely to be among the strongest in the area. You don't typically expect to use a distant WiFi hotspot, but rather one of the closest ones. And while a wall or two might be in the way, no hills, trees, or bodies of water.