Closed Loop O2 control and protecting against faulty sensors

Recently I replaced my ps1000 with an elite 2500 (the process was painless). I had just replaced my innovate mxtl's sensor with a new bosch LSU 4.2, and I was keen to see how well the long term fuel trimmed worked (I watched the long term fuel trim webinar which was very helpful - thanks guys!).

I have been having some difficulty getting it to work - but have posted that question in the webinars' thread. https://www.hpacademy.com/forum/webinar-questions/show/haltech-elite-self-learning-fuel-trims

Something I have already noticed - when we were setting the car up on the dyno, the tuner wanted to feed the lambda sensor directly into his software - We disconnected my sensor and plugged it into the line back to his controller. This was before we had turned off the O2 control.

We found the mixtures becoming really lean and decided to calibrate the sensor again to ensure it was correct - which it was. We then turned of the o2 control and found that the readings went back to normal.

What we learned - a zero vault signal will cause the haltech to pull fuel out as it believes the car is running too rich. I noted this down and would look into it further at home (this was monday)

On the way home from work yesterday, the mxtl started playing up and thus, the o2 control went stupid and started pulling bulk fuel out.

Obviously, this is bad, but i have found it's not uncommon for an O2 sensor to play up from time to time. Previously I never used closed loop control for this reason, but I am keen to see how well the haltech's new system actually works.

How do people protect against a faulty sensor? I have the following ideas:

The percentage enrichment/de-enrichment is small so even if reading a 0v, the amount of fuel removed will not be dangerous. Issues with this is it could screw with the long term fuel trim table.

Change the minimum and maximum raw values to 0.5v and 4.5v - that way the computer will ignore a 0v reading. This is how I assume the minimum and maximum raw values work under the diagnostics tab for the 02 sensor, but this is yet to be confirmed.

Anyone have another other ideas? Or have you experienced this as well?