>>You might think about checking another angle: from the above you have approx. 0.1% error on all SQL commands (16 errors on 600 * 25 SQL), so the probability of having the SQL error out again is at every thousandth error. Running the test for an hour should give you about 500 errors or a 50% chance of having the same SQL fail twice in a row. If you run your test for a few days on a dedicated machine you could make sure if the error always occured on the first SQL or if it occurs in the same percentage of "first errors". (Hypothesis: some trouble while creating/initializing).
>
>Yes, correct, when I ran more tests, at one occurence, it took 3 tries before the attempt was a success. But, again, on the user side, it was ok. I also dropped the number of occurences to 4 and increased the delay to 0.5 second.

Too bad - this probably means the error can hit you anytime. Running the few lines from OleAddapter creation over .Fill as a critical section might is the only thing left I can think of - and I doubt it could be that simple.

good luck

thomas