Well I can only say this- that I document all of my gas mileage (down to 0.1 MPG- not just weeks per tank) very thoroughly, and there was no appreciable change when I removed my O2 sensor. My car is a true daily driver, and I think I have over 50K miles on the unit.
Gas mileage changes slightly with seasons and weather, but nothing more (percentage wise) than any other car.
I am willing to bet that because I do actually drive so much, and spent so much time tuning, that my tune is one of the closest to "ideal" (relative to the car) that can be found. Many users do not spend the amount of time tuning that I have, and no matter who it is (Lance Ward included) you cannot tune every cell on the VE table in a single day to the extent which I have- this requires trips up long hills and down again, along the interstate at 60 MPH, 80 MPH, day(s) at the racetrack, around-town driving, driving while semi-warm, hot, in the rain, etc. Can you make progress in a single day? Absolutely. But it has to be a full day, dedicated to tuning.
Just because a car behaves when the O2 is closed-loop doesn't mean the tune is accurate. We have all read posts on this board about people who assume the WB will make up for all of their sins, and never bother tuning the VE tables properly. I am not saying this is the case here, but I want to know more about the tune before I disregarded this possibility (which would require driving the car).
I am curious to hear any theories why you have seen such dramatic changes in fuel mileage, with an accurate tune (which I would define as 5-6% max correction in any cell).
-Bob Cunningham