The ability to run sandboxed workers from a PostgreSQL-based queue sounds useful all by itself, particularly for multi-tenant SaaS applications that need to do compute-intensive tasks on untrusted inputs. The difference, of course, is that in this case, the worker implementation is part of the application, not a user-defined script. How tightly are the work queue and sandboxed worker runner coupled to the rest of the product?
They are tightly coupled at the moment but easy to extract out. I did not think there would be interest for that but I could carve it out as a separate dependency project for everyone to reuse!
We are thinking about creating something like this actually. We are still early-stage and haven't launched yet but we are looking for feedback and input to our idea.
