Soon (11.2) VNL will at least have bonds. We'll see how configurable we can make in the first iteration, but it is interesting to hear user input like this, how you want it to work.
JMol is very configurable, at least if you don't mind doing a bit of simple scripting in their macro language. I have even used it to make plots where the bonds have a radius proportional to the Mulliken overlap population (sort of a measure of the bond strength). You can easily choose which bonds should be displayed, and set the radius etc of each one individually.
The official JMol macro language is documented at
http://chemapps.stolaf.edu/jmol/docs/, but it's a bit hard to penetrate. The important thing is to check out the examples (drop-down at the top of the page).
I think we can do a better job at making it user-intuitive, their menu system etc is quite messy. But it will take us some time to implement all their functionality, they really can do a lot!
