I would like to point out that there can be several spice models for the same part, depending on what one is trying to simulate.
Simulating an opamp to verify ones schematic makes sense, checking functionality etc., requires a model that can be relatively simple. In many cases such a spice model van be downloaded from the manufacturer for free.
Simulating that same opamp to verify its behaviour in the schematic as part of an estimation for noise, or high temperatures or some other difficult special case at the limits, requires a different and much more sophisticated spice model.
In the simplest models, only the intended functionality is modelled, but the unwanted limitations of the part are not.
Who would have thought a TL082 opamp, under unfavourable conditions, can get stuck at one of the rails, for instance? This kind of (usually unwanted) behaviour is seldom modelled by free models.
There is usefulness in both types of models, because the simple model requires less calculations and hence can be simulated faster. In many cases, there is no need to simulate correct real life behaviour at the limits, but sometimes there is.
Imagine the maintenance on sophisticated spice models as the manufacturer, in time, goes through several versions of the same part. Imagine modelling the temperature behaviour of parts the are available in several different packages.
Do not not get me wrong: I would love to have easily accessible reliable spice models. I would absolutely love it! Even the simple versions of spice models for many parts would be very nice.
However, the complexity of the subject and the licence issues will make the effort quite hard to get completed. If only we would live in an ideal world where manufacturers could openly “publish” the limitations of their parts in spice models for all the world to know. I do not think we will see this any time soon. It is much to difficult to admit limitations of parts.