Serverless engine for long-running processes

Has anyone been looking into creating a serverless implementation of the BPM engine?
I am aware of the blog post describing how a short-lived process can be executed as a serverless function, but what I am looking for is an implementation that would support long-running, stateful processes, by outsourcing the async executor agenda to some external scheduling system.

The idea being that an engine only needs to be instantiated when a timer fires, or when a human comes in to work on a task - at which point the process state can be reloaded from the DB.

Any thoughts or pointers around this?

When we did the serverless article/youtube video (The Flowable Engine as a Serverless Function – Flowable Blog), there was not much ‘stateful’ yet on the market. Now, it seems like things have progressed and it looks like even regular databases can somehow be kept ‘ready’.

In theory, everything there should be applicable - the tricky thing will be getting the connection pool set up fast enough.