Frozen atoms are moving
This bug is mysterious for me because I can't isolate it in a minimal system, so I apologize for complex and messy example - I just can't reproduce it without whole production system.
Here DU particles are frozen dummy "anchors", which do not interact with anything else except one bonded SW atom. The group DU is frozen but the atoms still move slowly (see attached output from `gmx traj`). The distance between two layers of these dummies is decreasing with time which is disastrous for my setup. Expected behavior is fixed positions of all frozen particles regardless of any box changes.
Tested with gromacs 2019.2 and the reasonably fresh git version.