>I think this is related to SQL Server. Basically, if I go in my Hit table, I see that this first hit takes 5 seconds instead of 0.2 second. That is the time the hit took to be processed. So, it went immediately in the hit but took longer to execute. I will activate a SQL trace and see how it goes the next time. So, on that particular SQL Server, there could be a setting that forces it to go to sleep if no hit against SQL Server has been done within two minutes. Could that be?

Doesn't this bring us back to the SQL Connection pool issue already discussed? Bear in mind that there will be a connection pool for each connection configuration. But even if you were not using pooling the delay seems excessive.

Do you have your own instance of SQL Server? If not then other sites using that instance may be affecting the workload....
Maybe it's just an intermitent networking problem?