Future
Love the idea of this, the current nature of the official runner creating a Pod
rather than a more retryable resource like a Job
always seemed strange to me. I want to execute a thing until it either fails or succeeds, system failures should be the runners responsibility to solve, your solution seems to adopt a better approach. What is its future?
We heavily use services
(for https://github.com/bsycorp/kind) so using it in its current state won't work for us, would it be worth looking at a PR to add it? Or is this essentially feature complete for you?