MicrosoftThat's great, but what about recovering from failure cases?
Higher-level functions are nice, and moving state from explicitly-controlled systems into the FaaS-controlled storage is nice, but when I do that I also lose control. Right now, I can inspect, modify, and correct all queues, intermediate states, distributed leases, etc. in order to recover from bugs with full fidelity (absolutely no data loss). How would I do that when all the intermediate state is outside my system?
I think what Durable Functions is missing is all the documentation that's needed by third-shift on-call people who need to fix everything after it's gone sideways.
