Hi Thomas,

> I don't see deadlocks with the following test case.

Thank you very much for the test case!

> No gaps, no deadlock messages, no pk/unique violation in the log table etc ...

Same for me. Got deadlock if try with NO RECORD VERSION.

> It is crucial that accessing the central generator table is tunneled through a single access point. The SP in our case. Any chance that anything else is accessing the table in your scenario resulting in the deadlock / write update conflict?

I hope no. The user can modify the properties of the generator but just
if the generator value is 0.

Gabor