The software has limits on the mixture range, if it can't achieve O2 cross over within this range it creates and error and goes into a limp mode which is the same as if the O2 sensor is detectably dead. Stranegly this limp defaults to 6% CO on the B230FT, which really isn't necessary, but thats beside the point. It can take half an hour for it to decide the lambda is not right and light up the light, 6% CO is the max overfueling it can goto.

Once it's in Limp Mode maybe but before it gets that far (remember it takes at least half an hour or more of constant running to determine the sensor is dead) it will still try to overfuel. I once had a Lambda sensor fail during the emissions test and the CO was steadily climibing - the tester pulled the probe when it got above 9% but in fairness that wasn't a redblock. However it's fair to assume that until it finally decides the sensor is dead it will continue to increase the fueling and on the car i mention above, it took 3/4 hour to determine a dead sensor.

True, its certainly gong to depend on software, but from a safety point of view, it needs to be safe in a single fault condition and I think it would be unsafe and rather pointless to continually increase mixture until a potentially errant O2 sensor responds, hence the limits.

It seems that the limits in LH2.4 are more cautious.