Destroy panic error handling
If runner can not be destroyed by tf the runner could send back a comment to the repo with a link to the machine to be disposed.
Your machine could not be removed!!! You have to do it manually to prevent further charges.
perhaps some c/p actions/(gitlab -eq) that users can run a schedule is a better solution?
- https://github.com/iterative/cml/issues/973#issuecomment-1146149999
- https://github.com/dacbd/gcpcc#usage
- discord#cml/984180090114019368
my 2 cents is that this is probably p1 feature - notifying when a machine stays up is an important failure. doing "GC" on the user/app level is great architecture suggestion, but its out of the cml tool's control. On the tool level,. it needs to be "responsible" and "friendly" with such severe failures. so imo -comment on gh is must, and maybe we can add slack alert or similar?