There's another problem with closed loop systems, and that is the unreliability of using O2 content derived "A/F ratios" in determining whether the correct amount of fuel is being metered in the first place.
The idea is that under perfect conditions (i.e. 100% efficiency), there is some amount of oxygen required to burn a known quantity of fuel with a minimum of leftovers. Unfortunately, that efficiency varies quite a bit. There must be some "average" efficiency then that is used to calculate "A/F ratio" from oxygen content. So, we've arrived at the first problem: There is no direct correspondence between oxygen (or any other gas) content and "A/F ratio". Second, the ideal "A/F ratio" varies from engine to engine and from one rpm/throttle position/load combination to another. Third, I'd like to consider what the goal of tuning is in the first place. Is it to get a particular "A/F ratio"? It's more likely that we would like the engine to make horsepower and be efficient (i.e. get good mileage). Looking at oxygen content has limited use when trying to achieve these ends. However, don't get me wrong, oxygen measurements are actually good for something. Once you have tuned for max power by adding and subtracting fuel, the oxygen numbers can help you to detect an ignition timing or stagger issue. If you go straight to a particular oxygen content, you completely ignore these factors. Even once you've corrected these, the oxygen content can still vary substantially. So it can help to point you in the right direction, but isn't the final arbiter of anything. CO on the other hand can, once a proper value has been established through bracketing, be used throughout a map as a target to set the mixture strength close to optimum for power and mileage. CO is also not the final arbiter of anything, but it tells you much, much more about whether an engines map is in the ballpark than O2. You can hit an oxygen target right off the bat while the engine has the completely wrong amount of fuel and/or wrong timing, whereas with CO, fuel will be close, guaranteed. As an example, there are spots on two Futura dyno charts I have, where the O2 content is 0.2%. One has a CO of 12.6% (this is drowning rich!) and the other has a CO of 3.2% (perfect at this particular throttle position/rpm combination). I'm sure you can draw your own conclusions from this example.
The problem with CO for closed loop injection is response time, and I'm not sure if it can ever be improved enough to be useful.