I suspect it's simpler than this. When the car is at a steady state speed, it requires x horsepower to maintain that speed. Tesla sets a buffer over and above x for hills, passing, and general responsiveness. When the front motor alone can produce x+y horsepower, it sleeps the rear motor. When it can't, it runs both.
OK, but we're kind of saying the same thing.
It takes more horsepower to maintain the same steady speed on flat ground if the temperature is colder. So it's possible that simply due to the cold, the limitations of the front motor, and the value Tesla is using for y, torque sleep isn't happening at all at highway speeds in the cold.
And if that is true, Tesla should have anticipated this issue, and informed us about it. There's nothing in the JB Straubel blog post that indicates that torque sleep won't work if it's cold.
This would be kind of a major issue for those of us worried about range, living in cold climates. Because we had the initial range estimates, then we had the whole reduced range issue, and that was largely resolved by the blog post and the promise of torque sleep. But it is the winter when I'm most worried about range, because I know that's when my range is going to be the shortest. If I'm not getting the benefit of torque sleep whenever it's cold, that's a pretty significant change.