Hilmar,

>Update: I left the program running; using VFP 6, the program crashed at i = 64350, i.e., that many properties were added. Pressumably, a few of the built-in properties summed up to a round number (like 2^16, or 2^16-1).

AFAIR that is a NTI border line, the same as in the "normal" variable space of vfp, which is reflected in every object again.
>
>But things got very slow, a long time before the program crashed.

I think the slowness is at least partially the fault of the version[vfp6] - there were quite a few different runtime characteristic changes between vfp6 and vfp8 (I skipped vfp7). Parts of this is described in Calvin Hsia's blog.

Regards

thomas