Hi Philippe,
The Event Server is NOT a very resource intensive application. In fact, our sizing guide doesn't even recommend allocating CPU or memory aside specifically for it. It doesn't do anything but monitor disks for file changes so it should never consume a large amount of CPU or memory.
You may want to use Perf Mon to monitor the CPU and RAM usage. If you see it spike, then this would indicate a bigger issue to me, Please refer the below link for process monitor tool available in SCN wikipedia
3rd Party Tools - Business Intelligence (BusinessObjects) - SCN Wiki
The next step would likely be logging/tracing them to see if the logs have any relevant info in them. As for # of events each one can handle, we have no guidance on this that I could find but personally, I would say that this number is more like 100+ rather than less than 10 like you have.
Perhaps it's the shear number of Event Servers that is causing the issue for you. The CMS has to poll each one to check availability and get a list of events that it monitors so maybe that is causing the issue for you. Can you try taking it down to 1 per server to see if that helps the situation at all?
Also, there were some known issues around the CMS restarting without the Event Server being cycled. Is it possible that the CMS is going down sometime before these issues arise?
Thanks,
Daya