I don’t think making SCOM a manager of managers is a good idea. A tool like Splunk or an event management tool like NetCool is better suited to that task. Of course those solutions cost $$$ and you mentioned you were trying to keep costs low, but event management is not something you should skimp on.
It’s been too long to find the answer for my question with regards to network device monitoring limits.
Microsoft limits are based on number of monitored ports, with about 10 ports per device on average being monitored. Which ports are monitored is automatically selected by what is connected to the ports unless you override the defaults to monitor ports that are not connected to monitored computers or devices, or to disable monitoring of ports that are stitched to other ports and would be automatically monitored otherwise.
The fewer interfaces you monitor, the more network devices your management group can support. For example, to monitor 6,000 devices in one management group, you would need to monitor no more than about 3-4 ports per device, which might require you to disable a lot of default monitoring. And you would need very powerful computers to serve in the monitoring pool because management server is the the one doing all the heavy jobs when it comes to SNMP based monitoring. And recall that your Operational and Data warehouse databases are shared by all management servers and can be a bottleneck too depending on the I/O of the SQL servers.