Okay so here's what happened. I put in some debug code to do the benchmarking and forgot to take it out. This debug code reported that ALL servers were up and running on the development box when in fact they weren't.
On our development, in order to test worst case scenarios, we have a connection to a laptop that is running a VirtualBox install of CentOS with OpenVZ that is on a standard home ISP. This computer was off.
So basically, I told it to send out 30 SSH calls to a server that was reported by the system as "up", when in fact it was down. This is why it was slower (12 seconds compared to 1.5 seconds).
We'll test on this setup once that computer is turned back on

.
So for now, it appears this feature will be GREATLY improved in the next release.