TL;DR: JOGL or LWJGL? I'd have to figure out a solution either way
The main problem of the client rewrite is rendering. There are 2 ways:
1. Dead end way - pick any OpenGL binding and reimplement the current rendering. The problem here is that to obtain the best performance you need to have a deep knowledge of the rendering pipeline and techniques. People spend decades of 40-hour/week paid job to master that skill and move game engines forward. You definitely don't want to do that
2. Impossible way - pick any mature game engine (Unity, Unreal) and use it to render the game. The problem here is that all game assets are packed in a very specific handcrafted format. To re-use models, animations, textures, and everything else you will need to invest time creating a resource converter. There were converters to/from a very simple obj model format (and I did one myself), but the obj format lacks complexity and "better" formats are hard to grasp. There is also a possibility that we simply do not have some crucial info required to convert res into a mature 3d format
Due to the reasons above, I would not pick Java for the game client. Even for the (1) approach, it is better to use some mature game engine (at least with some built-in UI toolkit). Really, there is no reason to use Java - the game client "logic", apart from the UI (which ideally should be reused from some game engine), took me 3k lines of code to implement with zero code reuse from the original client (I'm talking about networking, resource handling, message processing, remote UI, and maintaining the "game world state" visible to the client)