FastAPI async endpoint: which s3fs client should I use? Needed example

If I wrap the s3fs client inside a DAO class in order to create an interface to add features with the remote fs, which s3fs client should I use?

For instance, assuming I create an async client, because I wish to use either example_one or example_two "coroutines/methods" in FastAPI async endpoints, which example should I use?

class S3Manager:
    def __init__(self, bucket_name: str = None):
        self.bucket_name = bucket_name
        self.s3_fs = s3fs.S3FileSystem(
            anon=False,
            asynchronous=True,
        )
        
    async def example_one(self):
        session = await self.s3_fs.set_session()
        work = await self.s3_fs._glob(...)
        await session.close()
        return work
        
    def example_two(self):
        return s3_fs.glob(...)

Since FastAPI handles its own loop to manage async coroutines, I believe the best approach would be to create the async client, leveraging the same FastAPI loop somehow: e.g. delegating the management to it. But:

1. the async client and example seems cumbersome: it seems like the s3fs library wishes to user to exploit the sync interface/client which itself uses the async support behind.
2. is there a performance gap by using the sync blocking client by calling its methods in a async coroutine (fastapi)?
3. how should I setup a complete example using FastAPI own loop?

Start with the issue's S3Manager examples and the s3fs.S3FileSystem asynchronous=True setup, then compare the sync and async methods in a FastAPI async endpoint. Document which client pattern should be used, whether blocking calls affect the event loop, and provide a complete example if the project supports one.