What was the default value of compaction_throughput_mb_per_sec in Scylla 4.6 and did this default value change to 0 from Scylla 5.1 ?
This configuration item is not used by ScyllaDB, it exists only for backward compatibility with Apache Cassandra, its value is ignored.
ScyllaDB uses the user-space scheduling provided by seastar (the framework it is built on) and has controllers for compaction CPU and IO, making sure it uses the appropriate amount of these resources, based on the current compaction backlog.
See this blog post for more details: Taming the Beast: How ScyllaDB Leverages Control Theory to Keep Compactions Under Control - ScyllaDB.