Wanted to share this with the community: I work with explicit water systems and with charge transfer complexes. Often I face an issue that with default tolerance parameters I either results in local minimum which still does not give me a good guess for TS geometry or for charge transfer complexes system oscillates continuously (over 500 iterations) around the final structure.
With some recommendations I became familiar with a paper and found that switching from internal coordinates to translational and rotational for such systems actually leads to a converged geometry optimization faster and with more fine tolerance. Authors made their code geomeTRIC available and it smoothly runs on macOS calling QChem for calculations.
I actually had issues running it when installed with pip or python setup.py install, but if installed using conda, it works great.
Hope it helps!