There is some good work going on to improve this. Faster WiFi speeds mean that less time is needed to transmit the same number of bytes, meaning that stations are less likely to interfere. (This is why everyone hates your 802.11b-only client that can do a maximum of 11Mbps. It ties up the entire channel for no good reason.)
The default rate control built into most APs is awful, and makes things worse than they need to be. For example, let's say sending at 100Mbps, there's a 50% chance that the message won't be decodable, and sending at 1Mbps, there's a 99.999% chance that the message will get through. APs will try sending at the high rate, notice an error, and then go to the low rate. But that's wrong: it doesn't minimize the use of airtime. Sending 100Mb at 100Mbps twice is much more time-efficient than sending it at 1Mbps once. Some work is under way to improve this algorithm and get it in consumer routers.
There is also work to make APs transmit to stations at a transmit power tuned to be as low as possible. If a station is close, the AP only needs a small amount of power to send a decipherable message to the station. And, because the power is low, it interferes with less space. I believe this is now working under Linux with ath9k chipsets.
Anyway, "channels" is a terrible way to understand WiFi performance, because space and time are the other critical dimensions. Unfortunately, most "WiFi analyzers" only show which channels have APs, which is a meaningless metric.
Here's an interesting talk about how this all works in the real world: http://apenwarr.ca/diary/wifi-data-apenwarr-201602.pdf