Indecision truly is worse than wrong decision. After waffling a bit I decided that the easiest thing would be to just close the storage before fork, that pretty much eliminates the weird stuff. Unfortunately, it does NOT necessarily fix the issue from #924 where the user was interacting with storage (presumably from a thread?) pre-fork as an import-time side-effect. For that, I think I'll just suggest working around it in your code-base or don't run that stuff as an import-time side-effect.
- Close the storage connection before forking process workers, this just simplifies things. Refs #924 sorta.
- Run the callback of an empty chord with an empty results list. Previously it was never enqueued, fixes #926.
