I agree with SAP to restart the servers at least weekly once or monthly once to resolve the stuck processes, orphans or zombies in the system.
I have seen when the CMS db connections overload with SQL server happened it didn't resolve the issue with simple SIA restart then it forced us to restart the servers.
Either the servers are VM or Physical it is best to restart the servers weekly once in the following order.
If you are in clustered mode assuming 2 CMS nodes with multiple processing nodes.
I would recommend to setup some batch job to execute all the stops/restarts from one server instead of from each server.
1) Write the batch script and place it on last processing node ( Assume 2 CMS nodes and 4 Processing nodes then script will run from 4th processing node)
Stop the SIA/Tomcat on Processing node 4
wait for 3 minutes
Stop the SIA/Tomcat on Processing Node 3
Wait for 3 minutes
Stop the SIA/Tomcat on processing Node 2
Wait for 3 minutes
Stop the SIA/Tomcat on Processing Node 1
Wait for 3 minutes
Stop the SIA/Tomcat on CMS node 2
Wait for 5 minutes
Stop the SIA/Tomcat on CMS Node 1
Wait for 5 minutes
Restart CMS Node 1
Wait for 7 to 10 minutes
Restart CMS Node 2
Wait for 7 to 10 minutes
Restart Processing Node 1
Wait for 7 to 10 minutes
Restart Processing Node 2
Wait for 7 to 10 minutes
Restart Processing Node 3
Wait for 7 to 10 minutes
Restart Processing Node 4
You have to make sure that your CMS nodes are starting prior to processing nodes to avoid any issues.
Make sure no network communication issues between servers other wise some times it leads to some services on processing nodes lead to either Running/Disabled or Initializing/Disabled which requires force termination and start of the service.
You have to do the validation testing after every week restart.