Asynchronous Downloads for Large Reports in Superset

[Bhav2308]

Using Superset v4.1.2 (deployed via helm in Kubernetes).

What is the recommended approach to enable fully asynchronous downloads (e.g. CSV, PDF) for large datasets in Apache Superset?

Currently, superset exports seem to be synchronous and can cause timeouts or performance issues for large datasets. We have seen superset crashing or restarting when it comes to handle large queries (50 million records)
We want to enable users to export large datasets without any performance impacts. To what extent Superset supports async queries?
We want exports (especially CSV's) to be downloadable via asynchronous mode. Is there a scalable method to handle this?

[jklaise]

Whilst Superset supports async query execution via Celery workers, downloading the results still load the whole set into memory on the server, see my comment here: #33243 (comment)