Derek Ortt wrote:then you should understand how errors grow, not decay, in time. If a model has a 200NM short term error and is good after 5-7 days, it has done something right for totally the wrong reason
While errors in general do tend to grow with time, that doesn't necessarily mean that the model is doing something right for the wrong reason. For example, a model may initialize a storm 200nm off in the beginning, but it is entirely plausible that such a displacement could have no real effect on the track of the cyclone. (I.E., a strong hurricane under a strong and expansive ridge). The initial placement of the storm could be 200nm off, but that ridge would affect the storm the same way regardless of the initial error.
Think of a ball rolling down a hill. Whether you start the ball at the top of the hill, or half way up, it's still going to roll down to the bottom. So in 5-7 days the ball would be at the bottom of the hill regardless of where it started from.
I think your reasoning is faulty.