RMAN performs backup and restore operations through channels. Each channel is an independent server process that reads from the database (during backup) or writes to the database (during restore). With a single channel, operations are serialized: one datafile at a time, one I/O stream to storage.
Parallelism changes this. With multiple channels, RMAN distributes work across processes. During backup, different channels back up different datafiles simultaneously. During restore, different channels restore different datafiles in parallel. The result is aggregate throughput that scales with channel count, up to the limits of the infrastructure (CPU, network, storage).
FlashBlade as a backup target is rarely the bottleneck. The practical limits are typically database server CPU network bandwidth, primary database storage, or the number of datafiles available to parallelize.