Currently we use `git(1)` to figure out what commit an image (in `img`) is on. This can be slow when Git decides to do a directory scan. Can we avoid that? options include: * cache the commit in `/ch` somehow * adjust `git(1)` arguments * examine Git metadata directly One place where this appears is if you have a two-stage build where the first stage is large but unchanged. It will take a long time figuring out the first stage has nothing to do.