||Sorry for the delayed reply.
I have been researching the issue and am finding it hard to describe.
Say we have the ArcGIS Server setup with 1 Nonpooled service. We have this setup to accept say 10 instances.
Now the server will work away serving requests with no problem. There comes a time when the services will stop responding and if you are using a web developed application this will return a timeout error.
In general for every instance, or client using the ArcGIS Server, there is a corresponding ArcSOC process, and for every request made by that client, the ArcSOC process uses memory and CPU.
When the web developed application returns the timeout or i refer to the "process's locking up", I have noticed that inspecting the task manager, one of the ArcSOC processes will be using CPU, not so much memory, but it will continue to use the CPU (not heavily), and will never stop. If i end this process, suddenly the ArcGIS server starts serving up requests again and the web developed application is funcational again.
So, something is locking up one of the ArcSOC processes and if one of them locks up, this locks up any request to the ArcGIS Server, which in turn makes the application unuseable.
I have yet to find out what is causing the ArcSOC process to lock up as it is very hard to replicate, but does occur.
So I was wondering if anyone had come across this activity before?.