In the article, it states that after a certain point, parallelization hits limits. Is that due to Scylladb's architecture or other constraints within the system?

For reads (unless there is another bottleneck), by 10x'ing the parallelism, an application can compensate for 10x the average latency and still deliver the same throughput.

Scylladb talked about their IO scheduler a good deal:

https://www.scylladb.com/2021/04/06/scyllas-new-io-scheduler...

https://www.scylladb.com/2022/08/03/implementing-a-new-io-sc...

Naively, it seems like it should have been able to compensate for the Persistent Disk latency for read heavy workloads.