Currently, when restic rewrites packs as part of a purge, the process is single threaded.

In an Amazon S3 environment, this results in long prune times. For example, in our ~850k pack repository where we keep 7 daily backups from several machines, about 35k packs need to be rewritten each day. This takes approximately 3.5 hours with a single thread.

Parallelizing this would significantly decrease the amount of time required to complete a forget/purge cycle.