Restart failed workflows

[jensg-st]

Is your feature request related to a problem? Please describe.

At the moment a failed workflow is a failed workflow. When it is failed we still have the state data and we could potentially re-run it. If it is script-based e.g python, node etc., the script could be fixed and the workflow could restart from there. Of course the workflow can not be changed but the supporting data like variables or in 0.8.0 files / scripts from git.

[alankm]

Restart from there? As-in, restart part-way through the workflow? This may be harder than it looks. Or very limited in capability. Yes, I think we could replay an instance from its final state. But what about workflows that did some attempts at error handling? How many states back do we need to go? Intermediate state data isn't saved, so anything other than the very last state is currently not possible. There's a lot of nonsensical states to resume to as well: what if the last state was an event listener? I also think this would cause strange behaviour in things like tracing and metrics.

The instance input data is available as long as the record of the instance hasn't been deleted. And it's available by the API. The frontend shows it. It seems far simpler and safer to have a button on the failed instance screen that retriggers the workflow using the same input data. This could be made to work with changed workflows.