Personally I've only had them permanently installed, with the 0-1v output connected to the lambda box. I was also datalogging RPM, MAP, and the frequency valve duty cycle via an Innovate LMA-3 and could see how the fast warmup caused the lambda ecu to push the system lean, working against the control pressure regulators warm-up cycle. Within about 60 seconds, when a 1-wire narrowband sensor would normally start to wake up, everything is more or less as it should be.
By default the 2nd analog output mimics a 1v sensor and its sharp voltage switch at 14.7:1. I never did get around to trying a different, more gradual curve to see if the widely swinging duty cycle range could be slowed down & give a more stable idle. Something to consider if you're looking for an experiment