## Clean up --stat and --sizes/--file-sizes combinations?

## Add check flag to delete, not warn, unreachable packs

Mention that they can get there from interrupted backups too.

## Download concurrency flag

Sometimes people are on different, slower Internet.

## Guard against colliding with another snapshot in the same working dir

Track working dir in some global state somewhere? Shmem?

## Compress after the cache?

Would require a fair amount of rework, but would also be a big speedup in-cache...
Probably not worth it...

## Can we pipeline filter/unfilter?

Trouble is that breaks the nice abstraction that a filtered backend is just another backend.
It also has to be *behind* the cache, which is always unfiltered...
