I think it would be useful to supplement the `reinject` command with an automatic
mode which calculates the checksum of the source file and injects the file if it
is known to the repository (without the need to provide a destination filename).
In addition, this could be done recursively if the user provides a directory to
inject. All this can probably be done already with some plumbing, but a simple
`reinject --auto` (or `scour`, or `scavenge`, if you like) would be a nice addition.
Of course this would only work for the checksum backends.

Example use cases would be:

* Recovering data from lost+found easily
* Making use of old (pre-git-annex) archival volumes with useful files
  scattered among non-useful files
* Sneaker-netting files between disconnected git-annex repositories