Dear Dr. Kohlmeyer,
thank you very much for your comprehensive and insightful response. According to the hints you have provided I went through the following steps:
I have tried with the latest LAMMPS patch version, namely: LAMMPS-64bit-27May2021-MPI. For the record, I execute the script throught the command:
mpiexec -localonly 4 lmp_mpi -pk omp 4 -sf omp -in .\initial_script.in
Unfortunately, the problem persists (even for twofold shorter timestep than initially).
Onward, I have included in the script:
fix mm1 all momentum 1000000 linear 1 1 1 rescale
and run the simulation for 500 rigid bodies (x11, ergo 5500 “atoms”). Below I attach the results for the components of system’s COM speed (v_x, v_y, v_z), which were monitored throughout the simulation via:
variable vcmx equal vcm(all,x)
variable vcmy equal vcm(all,y)
variable vcmz equal vcm(all,z)
As one can see up to 1.2*10^7 Timestep there is no drift, afterwards “the ice cube syndrom starts to kick-in”. “Horizontal lines” are results of the fix mm1, however it is visible that after each “reset” the system builds up to the state from before the “reset” and “step-by-step” the whole system gradually drift away. Frankly, I am in a blind spot.