[lammps-users] NVT calculation for point dipoles

But also in this case the tot. energy fluctuates strongly and then at some
point grows.

Can it be, that s.th. with the energy calculation is not correct for the
dipoles. Besides this, why is the rotational energy not added to the total
energy, or is it included in the pair energy or temperature?

It's not included b/c you didn't include it. You need to define a compute
that calculates the rotational energy and add it to the energy you
are trying to monitor. LAMMPS doesn't do this for you.
Look at compute erotate/sphere and compute temp/sphere.

Dipoles also require long cutoffs since the energy/force falls off
slowly. It may be that you will never be perfectly stable with
a cutoff, just as you wouldn't be for Coulombic forces.