NEMD zero temperature in source and sink region in heat flux

Hello,
I am using the NEMD method to measure the thermal conductivity of the 2D mos2 bilayer system. The two layers have a relative twists (moire structure). I am using SW and KC forcefield, as mentioned in published papers ( [Phys. Rev. Lett. 121, 266401 (2018)], [J. Phys. Chem. C 2019, 123, 15, 9770–9778]), and the moire structure is generated using the TWISTER code. The structure and forcefield work perfectly in NVT and NVE ensembles up to 600K. I am trying to measure thermal conductivity at 300K.

I use fix heat in NVE to dump some energy (0.0001 ev) in one strip of the structure and remove the same amount of energy from the other strip of the same dimension. I monitor the temperature of the entire system, and the two strips in thermo output. Initially, the output looks fine. Eventually, the temperature in both the strips falls to zero while the temperature of the entire system stays near 300K. The simulation continues in this stage for some time and then fails, stating that no atoms exist in the fixed regions. To see what was going on with the structure, I dumped the structure to visualize in OVITO, and it looks fine during the state where the temperature of the sink and the source region is zero. How long the simulation continues in this stage depends on how big the dumped energy’s value is.

In the past, I used the NEMD method to measure the thermal conductivity of 2D mos2, but I used only the SW forcefield. Is this problem happening due to the use of two forcefields?