The law of diminishing returns says that the deterrence effect of the 1000th day in prison will be far less than the deterrence effect of the 1st day in prison. In other words, lengthening sentences will reduce crime, but by less and less per day-length of sentence.
This discussion started out with someone suggesting the current sentences were TOO HARSH and should be reduced, say, with a warning / education-session on the first effect. This was justified by citing evidence that harsh sentences do not have that much deterrence effect compared to more lenient ones.
However, I would describe the current sentences as LENIANT, which means that they should currently be having a fairly LARGE deterrence effect. Agree that increasing them from here would have less and less of an effect.
The key here is that there are a lot of people who are NOT drinking and driving BECAUSE of the deterrence effect of the current laws. We are never going to get it down to zero. Perhaps we have cut it by 80% now, and cutting the remaining 20% gets increasingly more difficult. You could make it capital punishment and a stupid/addicted few will still drive drunk.
On the other hand, reducing the current sentence would likely result in LOTS more people driving drunk. Just because making it harsh wouldn't help a whole lot does not mean that the current sentence isn't working!!
From an incapacitation point of view a reasonable measure might be to seize the car of anyone caught driving drunk, even on the first offense. Also suspend their license. Permanentaly. No need to put them behind bars, taking away their cars and their license is a fairly harsh sentence that also has maximum incpacitation effect and a large deterrence effect with NEGATIVE cost (sell the car to recoup costs of running the RIDE program).