I'm not too sure it isn't that important, since the greater accuracy of such methods could potentially also allow you to use greater step-sizes. And thereby make the computation more efficient. Though obviously... I do not really know how much it will or will not matter. Regardless, I think I remember reading a paper some time ago, that allowed stepsizes of some 50days for simulating jupiters orbit along milions of years with good accuracy. I'll try to find it. Hmm... seems like it was a multistep method, which complicates matters when you want timewarping. (and would make it need another method to initialise values after changing anything) In any case here it is: http://adsabs.harvard.edu/cgi-bin/bib_query?1990AJ....100.1694Q If nothing else, it's interesting reading matter. Also I think the reason the keplerian orbits do not display the L4 and 5 points is because the the major objects don't revolve around the barycenters. (eg the mun revolves around kerbal which is in its own orbit, but kerbal doesn't orbit around the centre of mass of the mun/kerbal system) I'd expect L1 trough 3 to be more tolerant of this inaccuracy, but well... they're unstable.