Another poster-thread today a guy reported that he spent over $1000 on parts and inspections in California trying to his NOx smog down from huge numbers. He was about to give up. Then replacing the new NTK O2 sensor with a new Bosch O2 sensor dropped his NOx 98%
That reminded me that years ago I read about how to test (at least in part) an O2 sensor on the bench with a torch to drop the O2 level and raise it exposing the sensor to the flame. And using a volt meter to read it. I think the HO sensors produce their own voltage, and the Renix sensors have a changing resistance. They guy on the how too was using it to test multiple junk yard sensors and in a shop environment for servicing vehicles....
The NOx problem story is one where the sensor was not throwing an OBD-I code, but had been biased (something no ECU/PCM can test for with just one O2 sensor), by biased (signal bias error I think is the proper phrase?) I mean it was switching, making the ECU think the ECU was operating at Stoich, but was in fact making the engine run way too lean, and telling the engine everything was OK.
Imagine the sensor things it as at stoich, but is faulty, and is reading a lean condition, but reports the condition as normal, running at stoich, and the sensor detects and reports changes in the O2, but reads say 100PPM of O2 as 200 PPM of O2, that is a 100 PPM bias error.
Short of having a highly calibrated test stand, or a sniffer smog rig, I know of no way to test for that by testing just the sensor.