I get it. So, it would be a good idea perhaps to reduce damping effects
on DOF "i" if vel[i] is close to 0 (perhaps using some sort of
non-dimensional proportionality factor)?
It also suggests applying damping separately on each DOF (not like in
Hentz) is better, regarding this "feature", since the discontinuity
affects only one component in that case (or at least not always all
three of them).
