Databricks Autoloader is a popular mechanism for ingesting data/files from cloud storage into Delta; for a very high throughput source, what are the best practices to be following while scaling up an autoloader based pipeline to the tune of millions of events per minute;

While looking at tuning "cloudFiles.fetchParallelism" is something to look at are they any other configurations that need tuning? presumably fetch rate increase should be paired with delete rate from sqs/aqs as well ?