Iris model can't "remember" and retain fixes for more than a second or so with pan/zoom

[Editman]

I don’t know if this is a bug or a suggestion, since I don’t know what the intent was when building your various models…
I am using Iris-Medium, and I noticed that when the model has corrected and cleared up a word printed in roman characters, if the camera pulls back, that text starts to look more like Thai characters until it is unreadable.
Can’t the model(s) be taught to recognize dolly/zoom movements, and “remember” what it has repaired while close-up, then scale those fixes in lockstep with zooms/pans in the image unless something drastically changes the source pixels? It would seem that would be a helpful change, unless it just can’t be made to look good. What I captured below seems a bit silly for it to be that way.