fatal errors shut down the console

[stelcheck]

Instead, we should:

1. Stop the worker
2. Wait for file save to attempt restart

cc @ronkorving

[stelcheck]

@ronkorving would love to hear from you how I could do this. Basically, I am looking for a way to have the master suspend restarts on crashes.

Note to myself: this would also require moving the file watch logic in the master process.

[stelcheck]

@ronkorving any suggestions?

[ronkorving]

Not yet, sorry. I'm not forgetting, but this is a tricky one so need to really set aside some time for this.

[stelcheck]

I took some time to look into it. In short, I doubt we could build anything without at least exposing the WorkerManager object instance in the processManager of MAGE.

I can see two approaches which would work:

1. Block process shutdown of the worker, allowing for post-mortem investigation; the worker would then complete the shutdown once a file would be altered (which I think would be a neat feature).
2. Let the master catch the failure and control the restart logic.

Let me know if you think this analysis happens to be correct, or feel free to add your own analysis.