>>The record locking still works, but it is the caching that isn't working (at high load) that causes the problem, it just dumps the buffer (with no error) and continues on. From previous interactions with Microsoft, they are not willing to fix the problem as it will effect their benchmarks.
>
>I wouldn't call that "works". It worked, but then they added caching and voila, it doesn't quite work anymore.
>
>That used to be rock solid part of the code.

I wonder why there hasn't some litigation happy lawyer surfaced,
as the promised compatibility with the OS clearly has not been reached...