I don't think there is any writing on this because the details of this will vary by the techniques used (e.g. real geometry, displacement, bump mapping, etc.), and the render engine used. In theory, or rather, ideally speaking, your render engine's Level of Detail optimizations should automatically handle the reduction of detail in the background and optimize the balance between necessary render time and detail. Essentially, you're aiming for detail that is roughly pixel-sized to be fully represented, and anything smaller than that gets ignored or at least represented less accurately (and with less time consumed to render). So yes, it depends on render resolution.
But again the ideal is for the render engine to generally handle this. Not to say that this is perfectly implemented in TG or any other render engine, but you'll note Frank's comment above that trying to control displacement by distance to reduce render time actually did not work particularly better than TG's built-in (and, more importantly, automatic) LoD systems. So this is where things are working right. They don't always work ideally and sometimes workarounds are needed, but they are quite honestly almost always based around the same idea (control level of detail by distance from camera/size in screen space), and the actual requirements will really tend to vary scene-to-scene. There may be some basic rules of thumb you could come up with, but they will likely only hold true on scenes that are similar to the one you create them from, so you could maybe end up with a whole bunch of different rules or guidelines for different types of scenes, but ultimately it may be better to just develop an intuitive understanding of what elements do increase render time, and how to control their appearance at distance from the camera.
- Oshyan