>OK. I ran a test hitting a web service using Net 2 on a W2003 server with IIS6.
>Stress testing with just 1 thread on the client worked OK.
>Upping the number of threads produced the errors you are seeing. I ran ten threads and the server reported an average of 7.3 concurrent requests. Over 2 minutes and a total of 2070 hits I got 32 errors. The stress test was hitting the same web service twice with different parameters so the figures above relate to just one one these - overall the server was handling an average of 63 hits/sec.

Wow! Thanks, for a while I thought I had some badluck that only me what having that. It is good to know, well sort of, that this can be replicated for you as well. Based on your observations, the common factor here seems to be .NET 2.0. Yes, I had .NET 2.0 on all platforms I tested.

>I quickly tried (on the W2003 system):
>(a) Wrapping the web service in an aspx page with ASPCompat.
>(b) putting a lock around the web service
>but in both cases I still got errors.

Yes, this is exactly what I tried yesterday. The result were the same because .NET 2.0 and IIS were part of the same environment even with the Web Service.

>I can't help thinking that there's something simple that we're missing.....
>If I get time I'll dig a bit more tomorrow - but the chances of finding the time at the moment are slim.....

The bug is present within the environment .NET 2.0, IIS and VFPOleDb. So, it can still be within one of those components. I still believe there is something that has to be looked for from the Microsoft team in the VFPOleDb. I understand you were not able to simulate it under .NET 1.1. But, this could still be related to the provider.