Kyle,

I certainly wouldn't argue against such functionality, but with a Langevin integrator, it should not be necessary to thermostat different groups independently since there will *always* be good thermalization of all degrees of freedom if the friction is high enough. Multiple thermostat groups are something that evolved mainly to cover up problems with energy conservation that are not completely masked by weak-coupling thermostats (like Berendsen) where scaling the whole system kinetic energy would often end up with inequal partitioning of energy between two components due to poor transfer of thermal energy between them. Langevin pretty much guarantees this won't happen.