@merlin3650
I had some measurements a little further into the thread.  
https://bitcointalk.org/index.php?topic=5053833.msg47766551#msg47766551
I'll bring some of those numbers back to the first post in a moment or two.


@evade_57
For each device, cgminer spins up a dedicated set of supporting threads.  So they are suppose to work like their own instances.

In general, it's okay to mix until you run out of share resources, (cpu, memory, bandwidth, power).

A miner that doesn't behave can potentially draw resources away from healthy ones which:
- can affect other miners sharing the same regulator.
- can affect other miners sharing the same cgminer instance.

Probably good to group them in a meaningful way to control the collateral devices that are effected, should you need to:
- troubleshoot / tune a miner.
- restart cgminer.
- move the hub.
- change pools.
- restart the computer.