Atoms collapse while free energy simulation

GROMACS version: 2020.2
GROMACS modification: No

Hello, dear collegues!
I am trying to set up a free energy calculation for the following reaction of metal coordination complex hydrolysis (which occurs in a protein environment):
Me(S-Cys)3(OH2)2 → Me(S-Cys)2(OH2)(OH) + HS-Cys
I managed to parametrize reactant and product and created corresponding topologies, which demonstrate nice behavior in dynamics. Then I created dual-topology file which is accepted by gmx grompp and gmx mdrun. Unfortunately, ± on the half of the way simulation crashes,
and on the trajectory I see that upon breakage the Me-S and HO-H bonds S is collapsing with Me, and H is collapsing with O. It remains even after setting the charges of corresponding atoms to 0. What can be the reason?
I am not also pretty sure that my setup is appropriate for this task, as far as I could not find any option to break the bonds besides setting non-zero force constant for corresponding atom types in topology A and zero - for topology B. In this case bonds still “exists”, so this atoms cannot be paired for non-bonded interactions. 1-4 pairs also are not updated, etc. Maybe someone can propose another way for topology building.

As far as new users cannot upload files, I attach drive link to a folder with topology and structure FreeEnergyProblem – Google Disk

Thanks in advance!