Define typical problems users recover from
While users' systems can get broken in a multitude of ways, there should be some uneven distribution of problems. We should cover the most typical ones, especially those that don't arise from the user trying to be clever and failing. Once we define what we need, we can start implementing automated solutions.
What comes to mind:
- reinstall a kernel after the battery runs out during installation
- reinstall a kernel after space runs out during installation
From a developer's POV:
- reinstall a kernel after installing a broken one