I do pretty much the same thing in both LD4DStudio and LDCad, the biggest downside is deciding what's a part or not resulting in the code trying to flatten a whole model in some cases 
In the end I've put an vertex count limit on the prep function.
@Travis:
I've been using 32 bit indices for years, and as far I know this hasn't caused any problems for users at all.
Are you thinking about shifting LDView to a per part approach? Performance will almost certainly improve.
ps: Don't underestimate the VBO buffer changes, I've made changes concerning this in my LDCad dev code and it resulted in a performance gain of over 35% or even higher in certain circumstances.
In the end I've put an vertex count limit on the prep function.@Travis:
I've been using 32 bit indices for years, and as far I know this hasn't caused any problems for users at all.
Are you thinking about shifting LDView to a per part approach? Performance will almost certainly improve.
ps: Don't underestimate the VBO buffer changes, I've made changes concerning this in my LDCad dev code and it resulted in a performance gain of over 35% or even higher in certain circumstances.