AFAIK there is something called near clip (or some similar name) -- which is a distance value representing the nearest the viewer can get to something. Whatever that value is, there are an additional number of increments that follow (maybe 32k) with each increment representing a further-away distance used as a place to "drop in" other objects that are further away. It's a bit like a queue. How far away the last increment falls is based on how large a distance value you've assigned to the near clip. If you set near clip to see something like tiny, like 0.01mm, you're going to run out of increments before you get very far away because 0.01mm is so very small. That may be a perfect distance for viewing jewelry but not for GIS data. Setting nearclip to something a bit larger, say 1mm, will set the last increment much further away; Setting near clip to something large, like 1m, results in the last increment being very far away indeed.
Clipping occurs when you zoom inside of the near clip value.
I expect SU uses a near clip value that gave them what they felt was the right balance between closeup and furthest distance that can be seen. Obviously some modelers, myself included, occasionally need to get in even closer to see something.
I dunno what the solution for that is... it might be a limit in Open-GL, it might be a limit in a 32bit program, or something else, such as letting the user pick his own near clip value as part of the model attributes. Whatever it is, it isn't a bug, it's a decision that tries to find the right balance of near/far for the largest number of users.