Half a mile should be enough for warm enough changeover this weather and system may be set to start on gas if reducer is already warm enough.
Could be that your reducer temp sensor works properly some of the time but other times could be reflecting hot reducer when it is cold (usually implying intermittent short circuit).
Lambdas will improve matters anyway. Even open loop fueling (such as shortly after starting the engine) may be inadvertently affected by faulty probes.
Not saying there's a problem with mixture while you're driving, just worth checking and would expect (any installer) to check calibration/mixture while driving particularly after your type of problems.
Edit - In my last post I forgot to mention that certain injector and/or LPG ECU issues can create an imbalance of injector flow to each cylinder, an imbalance between injector flow means some cylinders get rich mixture while other cylinders get lean mixture. As long as the difference in flow isn't enough to cause rich/lean cylinders to misfire, lambda readings can still flick properly. So, even flicking lambda isn't 100% confirmation that some cylinders are not lean while others are rich. To check injectors flow the same it is possible with some (most modern) LPG systems to switch individual engine cylinders to run on petrol while other cylinders are running on LPG, if the system is calibrated correctly and all LPG injectors flow the same then switching any individual cylinder back to petrol shouldn't change the pulse duration of the other injectors. Bit more scare-mongering maybe
but it isn't very unusual to find differences in injector flow rates, particularly with certain types of injectors when they're old/worn. Usually has to be a pretty bad case to cause VSR so don't worry! Older ECU's (like your Stef) are less likely to have the facility for switching individual cylinders back to petrol but on engines with multiple cylinder banks fed by multiple LPG injector rails at least differences in overall flow rates between injector rails can be performed by switching which rail fuels which cylinder bank... If injection pulse time per cylinder bank (noted in LPG software) depends on which rail is fitted to which cylinder bank then this probably points to the average flow of the injectors (3 in your case) on the rail giving the leaner mixture to be less than the average flow of the other rail (and vice/versa) but this still doesn't tell you if the leaner/richer or both rails have a problem injector or more. Less of a issue if all injectors on the same cylinder bank flow the same (even with a difference between average flow of rails) than if individual injectors on a rail flow differently but the only correct situation is if all injectors on both cylinder banks flow the same.
Full time LPG installer
Servicing / Diagnostics / Repairs to all systems / DIY conversion kits supplied with thorough tech support
2 miles A1, 8 miles M62, http://www.Lpgc.co.uk