Apparently, in my calculation, the orbits of mutually repelling particles with an inverse square law have negative eccentricity (hyperbolic orbits with e<-1).  If I make them attractive, I get positive eccentricity again.