>More bad news - I can get errors on NET1.1 as well as .NET2.0

Similar error percentage or a different level ?
IAC, that takes care of any adventurous ideas chaining different runtimes together for write statements <g>.
What happens if you use the older versions of vfpOleDB.Dll, especially the one from vfp8 ?

>- in my original test I hadn't jacked the thread count up to the same level....
is there a clear trend visible from the number of threads involved (and if so, equally strong in 1.1 vs 2.0) ?
As you may remember, one of the things I spouted off was to use the sys() for defining a critical section,
but that does not seem to be supported by VfpOLEDB :-(.

Any attempts to implement something similar via recordlocking through OleDB-SP will probably
kill scalability right off (if not errorprone as well<g>), but perhaps some "isolation" can
be implemented via .Net.

OTOH some input from MS probably can cut down to the core of the problem faster <g>.

thomas