I was playing recently with a binary system  
( the two red planets which orbit their commom barycenter ( red dot in the middle) , see picture ).  
In the sim the yellow test particle was orbiting the binary .  
 
From this orbiting partcle I detached another particle , the red  one .
It moved to the binary , got some sweeps , formed a star-like orbit ...
I guessed it would crash soon with one of the binaries or being ejected from the system .  
To my surprise however the particle kept forming its star like orbits , with a nice pattern , changing its orbit with every pass a little ..
But t kept going ...
 
I've seen analogue orbits earlier , but never saw such an orbit for a particle which was releaed at a random position , at random velocity , around a random generated binary .  
Funny .  
 
Whished I had saved the sim .  
Fortunately I made a screenshot .  

Quote from Tony on 10/07/14 at 17:55:03:

Well , I ran this application not with GSim , but with a simulator I wrote by my own.  
It lets detach a probe from an Orbiter at command input , so I'm not sure anymore when I hitted the detach button and the probe was created .  
I wanted to simulate the Rosetta's landing on 67P . In GSim this isn't easy as we have a rather complex object and complex gravity potential field.  
 
On your edit : The screenshot was made with the barycenter of the binary kept in the middle . So both  binary stars are in orbit around their COG.  
 
Maybe if I repeat the sim I can get the same result after let's say 100 tries , not sure about this .  
What I tried to say is that imo this orbit is a realistic orbit and not an "engineered ,well defined" orbit .