ray: High memory usage Environments

Im working with an environment that has a very high memory usage. This usually prevents any sort of Async Sampling, since copies of the environment are very expensive. Is there any example of working with a high memory usage environment for ASync Sampling?

My understanding is that each state needs to be hashable, so it could be that because of the large state space - Im running quickly out of memory on instances with 500GB+ of RAM

So, a good question might be - when sample_async or high-throughput architectures are used, what data is duplicated?