I have simulated a protein-ligand system using GROMACS 2020.6. Later, I extended the simulation and subjected it to production using GROMACS 2020.3. I didn’t encounter any error while executing the above steps. Will usage of different versions affect the accuracy of the trajectory?
If mdrun allows you to the extend the run by reading a checkpoint file, there are never accuracy issues. It might be good to check whether bugs were fixed between the two versions that might affect your results.