The above message is about a Windows-based PMC
Your assumtion is correct. At this time, a single Manager run will not take advantage of multiple CPU's. I recommend a soft limit of 20-25 systems per policy file with any number of policy files per manager run. Typically there is one Manager run per unique workload characterization file.
You may certainly process multiple Manager runs at the same time taking advantage of a SMP system. The process you outlined will work fine except only schedule 1 run per CPU. 30 nodes per policy should work but keep the analyze times down to under 1 hr if possible.
How many systems are you looking to process with 1 Manager? 1000?
Btw, I'm working to have this limitation removed so a single manager run can process all data for all domains with multiple .an files and proper use of system resources.