Just a quick update.
We did two things:
First, to increase hosting server time-out from 30 minutes to 60 minutes. It didn't solve the problem.
Second, to increase the persistence timeout on the netscaler to 60minutes. It seems resolved the problem.
Thanks.