Thanks everyone for this great feedback. I'm still pumping it into my document and rolling along. I'm blown away by the 8.7m edged model and and am happy to hear that you're still a fan of SketchUp while working at that size! Most users (as has been seen since that post) don't approach that size and get frustrated well before they get there. In your case, Jopsa2 it looks like avoiding materials for as long as possible is a win, and then patience as you allow SketchUp to render. I'd think Sandbox tools could benefit from Multi-Core if you allow that it still has to bake, and of course the same for output to video or raster.
The changing over time in this thread really illustrates that its a hard question to define and to answer. The tips for modeling smarter are further proof to that and in their own way define limitations as well. Every post has been helpful so far (even those that wander into dislike of SketchUp.
Right now I'm looking at 3 key areas that need definition to help new users:
File Limitations, these are things like file imports and exports, what you add to or take away from the process.
Usage Limitations, things that are generally handled with smarter modeling. You might call it the educational limitation as not everyone has these problems and many power users find a way to overcome them.
Functional Limitations, SketchUp just doesn't create circles without segments, it won't model well under 1mm, and it really doesn't like models that are miles in scope.
Some of these are constant problems that we're aware of and constantly trying to fix, some of these are the result of SketchUp being so useful that people want it to do more than it was ever intended to do, and some of it is new information that its hard to see in a QA lab. I'm hoping to build some test case files based on this feedback and may do a new post in the near future with those files to see about getting a performance test from users of varying computational power or skill sets.
Until that time, I'm still reading and still quite appreciative of this data. Thanks everyone for your feedback so far. (c: