I agree with Andrew that figuring out why the model has problems should be your first concern and only when you can’t make progress there would I consider using stacking to get some inferences even from a non-coverging model.
I have drafted a list of strategies to handle models with divergent transitions that also (partially) apply to the case of high R-hats Divergent transitions - a primer Would be cool if you checked it out and let us know if:
- You are able to understand the strategies given
- You are able to determine which of those can be applied to your case
- Some of them eventually let you move forward
I’ve previously had premature termination of various R computations result in weird warnings, so I wouldn’t worry about the warning much.
Best of luck with your model!