The proof of concept provides the ability to run Enduro as a single process running all components in a monolith-like way: api, workflow workers and activity workers. The ability to deploy workers separately would bring multiple advantages: redundancy, scalability, etc...

Things we'd need:

• New entry point such enduro-worker or enduro --services=worker,
• A distributed semaphore - the current implementation is local only,
• Distributed config, since it's all kept in a file now,
• Address locality limitations in Problem: it is not possible to submit batches of transfers #137 (once implemented).

Our main workflow is already using the session framework.