CSMA/CA and multichannel concept have been around for a not short while. It would be stupid that nobody has considered multichannel CSMA/CA.
Well, the reason that seems to have cause the current situation, where single-channel CSMA/CA is more well known, is that single-channel CSMA is used everywhere in WiFi hot spots, which is because common WiFi APs don't communicate at more than one channels at the same time or jump frequently among channels. And this is probably because common WiFi environment is static without much dynamic interference, such that there is no need to switch channel once a channel is picked during initialization. For home use, a mobile station wants to connect to an AP and doesn't waste power to check the AP and MS are on the same channel, which is another incentive that AP locks on to a channel and all mobile stations perform CSMA/CA to contend for access.
If we are put in a more dynamic ad hoc scenario, which WiFi is also designed to accommodate, multi-channel CSMA/CA with frequent channel vector exchange between transmitters and receivers would come to use. By the way, frequent channel status exchange between the transmitter and the receiver is necessary for multichannel CSMA/CA.
For centralized networks, it would be better if the AP stay at the same channel and use it to death rather than the AP hops around several channels to accommodate different mobile stations. More single-channel APs can always be installed as interfaces to the backhaul for increased population of users, example being several APs each working on a different channel (1, 6, 11).
For distributed ad hoc networks, maybe multichannel CSMA/CA is more appropriate, because first the high mobility results in dynamic interference, second the distributed form make communication pairs more flexible rather than determined to be a star centered at the AP, so a smaller size of network has more flexibility to change channel.