Most cane models only worry about going out 3 days for details. That's why past 3 days is hatched in the maps. Do you follow the 348 hour GFS model to plan your picnics and vacations?
I doubt that an ensemble that learns from past biases is valuable anymore when the statisticians are constantly tweaking the models and their biases change