Writing as both a SketchUp user (for ~8 years) and a professional software engineer (who has done a modest amount of 2D and 3D graphics programming during a 39-year career so far), I think it is clear that there is a mismatch between the implementations of SketchUp, the DAE exporter, the FBX converter, and Hitfilm regarding the semantics and on-disk encoding of a 3D model. Surely it is possible to precisely express the aspects of a SketchUp model that ultimately yield failure in Hitfilm.
One might hope that any random collection of model attributes (edges and faces, face orientation, group and/or component structure, images of various kinds, guide points and guide lines, dimensions, layers/tags, scenes, materials, etc.) as created in SketchUp would be exportable, convertible, and importable into Hitfilm and successfully yield a visual result similar to that originally seen in SketchUp. Clearly, it does not always work. Perhaps the failures are due in principle to unknown (to us) semantic incompatibilities in the notion of a 3D model between SketchUp, DAE, FBX, and Hitfilm. Or perhaps the failures are due in practice to bugs or limitations in one or more of these four pieces of software. A detailed study of the specifications of these pieces of software could answer whether semantic incompatibilities exist (but it’s quite possible that no public and detailed specification is available for one or more products). Lacking that, a campaign of experiments would be one way to determine in practice what aspects of SketchUp yield ultimate failure in Hitfilm.
Briefly on the subject of a “good” vs. “bad” model, this is largely subjective. A primary reason it is subject has been stated a few times in this topic already: a model’s quality largely depends on its original purpose. A model of a Tulip plant in bloom consisting of millions of edges and faces, with roots, stamens, pistils, petals, etc. arranged with high fidelity to a particular original live example, might be a fantastic research tool to a botanist but useless trash to a film set designer doing pre-visualization. Or, a model of a railroad boxcar that looks good but does not consist of “solid” objects might be useless to a person wanting to 3D print a scale-model boxcar but would be great to a film set designer for pre-vis.
I doubt that many of the aspects considered when assessing a 3D model’s “quality” are relevant to the SketchUp->Hitfilm process addressed in this topic. Some aspects might be, such whether face orientation is consistently inside/outside, or whether image textures are used in some atypical manner within SketchUp. Others aspects, such as whether the geometry is lightweight (simple) vs. heavyweight (complex) might not a process issue.